Glossary of Sensemaking
Terms
| A |
| abstracting | Creating a representation of the salient features of an information element. An information element could be any information-carrying artifact, including a document, a segment of a document, an annotation, or even a group of elements. The representation could take different forms, including a well-chosen name, a summary, or an object and attributes representation of features. In the context of sensemaking, sensemakers often search for cogent abstractions to guide them in simplifying and structuring information. See also summarizing and naming. |
| abstraction | An intensional representation of a class, typically a description. Abstractions always summarize, that is, they emphasize some properties and leave out other properties. Omitted properties may be particular to members of the class but are not salient to defining the class. |
| abstraction tuning | Abstraction tuning refers to the process of modifying or "tuning" an intensional description so as to alter the properties that are included. When a set of abstractions is defined over a universe of examples, tuning the abstractions amounts to reclustering the examples in order emphasize certain distinctions for a purpose. See also representational shift. |
| accountability in sensemaking | Sensemakers differ in their skills and knowledge. Accountability refers to the ability of a sensemaker or system to give an accounting of what information was used (or not used) in reaching conclusions or making summaries. It includes a record of judgements that may have been exercised in ranking one information source over another, the use of multiple sources, and the reliability of combining information from multiple sources. |
| active document | A document with interactive computational support. To illustrate the difference between an active document and a passive document, a spreadsheet printed on paper is passive whereas an interactive spreadsheet on a computer is an active document. In the context of sensemaking, a sensemaking report could be active in the sense of having active links to refer back to sources. It could also be active in that it shows different sets of conclusions depending on which set of assumptions the reader selects. See also resumable sensemaking, source-linked sensemaking and assumption-linked sensemaking. |
| affordance | An affordance refers to a physical property of something that influences how it can be used. For example, the affordances of paper include its properties for being viewed, it's light weight, and so on. The nature of a handle on a door determine how one open the door -- by pulling, or pushing, or twisting, and so on. |
| "aha!" feeling | A feeling experienced by sensemakers at a moment of insight. See also perceiving order and "Oh no!" feeling . |
| annotater | A system for adding annotations to a representation. Annotations can be organized as properties. Annotations for human usage are often text. In the context of sensemaking, annotations could be used for expressing information relevant to sensemaking, such as rankings of sources or explanations about using or qualifying source material. See also hermeneutics. |
| antinomy | A contradiction between two seemingly true statements. |
| anytime methods | Anytime methods (or algorithms) try to find the best solution that they can in a given limited time. These methods yield better or more optimal solutions if they are given more time. Some anytime methods work by iteratively improving a solution, simply quitting when they are out of time. Others select different approaches (more optimal but more time-consuming) to solving a problem according to the time budget that they are given. People who carry out sensemaking under severe time pressure tend to use some form of anytime method in their work. |
| assumption-linked sensemaking | In complex and forward-looking sensemaking tasks, it is often the case that the sensemaker needs to make assumptions. For example, these assumptions may relate to information that is missing or information about the future. In assumption-linked sensemaking, the conclusions in the sensemaking report are linked to representations of the assumptions in a way that supports assumption-based reasoning. For example, the report could be an active document that shows which conclusions depend on which assumptions. For another example, the sensemaking report may be source-linked or resumable, so that new information bearing on the assumptions can be easily integrated into the report and the report supporting updating of the report. |
| assumption linking | It is common practice in writing many kinds of reports to indicate important assumptions that underlie conclusions or an analysis.In the context of sensemaking, this is called assumption linking. If the sensemaking report is an active document and assumptions are represented explicitly, it is possible for a reader or sensemaker to go from elements of the report to underlying assumptions. If a sensemaking system supports assumption-based reasoning, then assumptions may be marked as believed or not in some belief set, and the display of the report can be adjusted to show which elements or conclusions are supported (or not) by the set of assumptions in the belief set. |
| attention management | In a complex information task, human time and attention is often the scarce resource. Attention management refers to elements in an information system that are intended to take into account the resources and cognitive constraints of the user. Such systems attempt to regulate the rate of information and the priority of various information according to context so as to help the user to be effective. |
| audit trail | In the context of sensemaking, a representation of the set of layered sources, assumptions, and annotations related to a conclusion. See accountability in sensemaking. |
| B |
| broadening retrieval | An information retrieval system uses a matching process to decide which documents (or other elements) match a query or are equivalent to an exemplar. Simple matching processes often miss retrieving relevant documents because of syntactic or vocabulary differences. For example, a query may be looking for "court cases involving the government" but a particular document may be about a "court case involving the state department" where the word "government" is not present in the document. Techniques that broaden retrieval use additional knowledge -- such as knowledge of synonyms and part-whole relations -- to increase the set of relevant documents that are identified and retrieved. |
| C |
| class/subclass relation | A relationship between terms where one term refers to a more abstract type than the other. For example, the term "dog" refers to a subclass of the "animal." Sometimes the higher class is referred to as the superclass. See hypernym and subclass relation. |
| cluster | In a set of elements, a cluster is a subset of elements that are relatively near each other and relatively far and separated from other elements. To "cluster" a set is to identify the sets of similar elements. For example, a set of documents could be clustered so that documents on the same topics are put in separate clusters -- identified by the presence of the same low frequency words. |
| cluster label or name | A name for a cluster. Typically, the name is intended to characterize the common features or basic abstraction for the cluster. Also see abstracting and naming. |
| complete ordering | An ordering is a relation which specifies the precedence for all possible pairs of members of a set, thus defining an exact order of the set. Equivalently, a complete ordering uniquely assigns an integer to each member of a set between 1 and the number of elements in the set. Contrast with partial ordering. |
| concept splitting | The process of taking a relatively complex concept and dividing or "splitting" it into two cases. Concept splitting often involves dividing a concept into two concepts such that one is a more abstract or general concept and the second is a specialization of it. |
| concordance | Alphabetical index of words in a book or in an author's works with the passages in which they occur. |
| corpus | A collection of documents. Typically, information retrieval is done relative to a given corpus. An open-ended corpus is one that is growing or expanding during a sensemaking activity. |
| cross product | The pairing all of the elements of one set with all of the elements of another set. A cross product of a set of m elements with a set of n elements yields a set of mxn elements. For example, one might have a set of technologies and a set of applications. A cross product is a representation of each technology with each application. Tables are often a useful way of visualizing a cross product, where elements of the first set are columns, elements of the second set are rows, and each table entry describes some important attribute of the product term. |
| cost structure | The distribution of costs among the elements of a process. The term "cost" can be used generally to refer not only to particular financial costs, but to the use of resources generally, especially time. Roughly, determining the cost structure of a process involves understanding the individual steps and resources that are consumed by the process, and characterizing how they are allocated and how they scale with the size of the task. Determining the cost structure is a useful step in any effort to make a task more efficient, because it helps to focus attention or technology on the places where it can be most effective. |
| D |
| E |
| emitter | A program that creates or "emits" a kind of output. For example, a program may emit XML documents or outlines. |
| encodon | An instantiated schema. In a sensemaking task, a sensemaker fills out templates or schemas to capture information. For example, he may fill out elements of a table or fill out forms. The filled out items are called encodons. See also instantiate schema. |
| external cognition | External cognition is a phrase referring to ways that people augment their normal cognitive processes with external aids, such as external writings, visualizations, and work spaces. External cognition is human or cognitive information processing that combines internal cognition with perception and manipulation of external representations of information. |
| equivalence | A way of saying that two things are treated the same according to a definition of equality. In sensemaking, notions of equivalence are important for governing retrieval based on assumed similarity of meaning. For example, a matcher may treat two words as equivalent if they are synonyms. The notion of equivalence can be applied to any kind of representation, including terms, phrases and subgraphs. Two phrases may be equivalent if they express the same relationships between words that are synonyms. See term-equivalence dictionary and subgraph equivalence. |
| extraction | The process of identifying and obtaining particular information from a representation of a greater amount of information. For example, one could use a parsing or recognition process to isolate a statement of a particular fact from a larger paragraph that contains many other facts. |
| F |
| finite state machine | Finite state machines (FSM's) are machines (generally computer programs) which proceed in clearly separate and discrete steps from one to another of a finite number of configurations or states. There is a well-known relationship between classes of language grammars and finite state machines, namely, that finite state machines are capable of recognizing regular grammars. Regular grammars describe the simplest languages (in Chomsky's types) and are suitable for linguistic tasks such as stemming, spell checking, and many other operations. |
| fluid user interface | User interfaces designed to take advantage of the new higher-resolution displays and monitors. For example, fluid features include animations that increase the apparent size of the font on the line or lines where the user is currently editing text, while maintaining small letters over most of the working area. For another example, on a computational whiteboard, a fluid user interface may gracefully move adjacent objects over little when the object that the user is working on requires more space. Fluid user interfaces may prove powerful in providing a facile sensemaking interface designed to enable a user to maintain a focus of attention on new and changing information. |
| format stripper | A program that removes formatting information and other non-content strings from text. This canbe a preliminary step to equivalence matching. |
| frame | A representation consisting of frames and slots. Frames, also known as units or objects, are organizing elements. Frames have names and are typically related to each other along specific kinds of relationships, such as generalization/specialization relations, part/whole relations, and other. Each frame has a set of slots, also known as attributes, fields, properties or variables. Each slot has a name ("slot name") and a value. |
| FSM | see Finite State Machine. |
| fusion | As in "data fusing" or information fusing. This means combining information from multiple sources. |
| G |
| goal shifting | Faced with the task of finding and organizing information, a sensemaker does not always start out knowing exactly what he or she wants. After following one particular line, it may become clear that the sensemaker was asking the wrong question and should pursue a different direction. This shift in direction and goal is called goal shifting. |
| grain | The level of detail in a representation. To decrease the grain of a set of information means to obtain or represent more-detailed information on the topic. To increase the grain is to move to a coarser representation. |
| H |
| HDAG | Hierarchical directed acyclic graph. A graph structure with nodes and links. In an HDAG, a node can itself contain a subgraph. A link into a composite node in an HDAG is equivalent to a link to every root node in the contained subgraph. A link out of a composite node in an HDAG is equivalent to a link out of every fringe node in the contained subgraph. |
| hermeneutics | The study of the methodological principles of interpretation. Historically, many ancient texts -- such as the Bible in Western religions and ancient Chinese texts including the writings of Sun Tzu -- have accumulated layers of commentary, which are published along with the work. In each layer, subsequent reviewers comment on the original text, the historical context of its writing, the nature of translations over time, and on the contexts of earlier interpreters. In the context of sensemaking, hermeneutics is relevant because there can be more than one way to interpret information and because hermeneutic practice places demands on annotation systems for successively revealing layers of commentary. |
| holonym | A concept that has another concept as a part. Door is a holonym of knob. Opposite of meronym. See also part/whole relation. |
| hypernym | A word whose meaning denotes a superordinate or superclass. Animal is a hypernym of dog. Opposite of hyponym. See also superclass and term-equivalence dictionary. |
| hyponym | A word whose meaning denotes a subordinate or subclass. Dog is a hyponym of animal. Opposite of hypernym. |
| hyponymy | The class/subclass relation. See subclass relation. |
| I |
| J |
| K |
| L |
| learning loop | A cyclic behavior pattern that appears in many sensemaking tasks in which a sensemaker first searches for a representation and then encodes information in the representation to reduce the cost of operations. In searching for a representation, the sensemaker often incrementally modifies a representation to capture important regularities. Typically, a sensemaking process iterates between a top-down process of instantiating representations and a bottom-up process of searching for better representations. |
| linking | Creating pointers from elements in a document or workspace to referenced items. Links can be simple descriptions such as text citations to referenced items or they can be computational objects such as hypertext links on a web page. Links can connect to different kinds of referenced items, including other portions of a document, other documents in a corpus, information on the web that is computed on demand, and meta-data. Links can be used to support other operations besides following the link. For example, links can be used to guide particular interpretations and visualizations, such as reasoning about validity of conclusions based on sources of different quality or dependence of sections on assumptions with varying degrees of belief. See also assumption linking and source linking. |
| M |
| matcher | A matcher is a program that determines whether two things (such as terms, phrases or subgraphs) are equivalent. A matcher can also create a table of correspondences which indicates which elements are equivalent to each other. |
| media clash | Information available to sensemakers can come in a variety of formats and media. It is often difficult to transfer information from one format to another, or to combine information from different formats or from different media. The phenomenon of increased cost for combining information from different formats or media is called media clash. |
| meronym | A concept that is part of another concept. Knob is a part of a door. Opposite of holonym. See also synecdoche and part/whole relation. |
| meronymy | The part/whole relation. See also part/whole relation. |
| metonymy | Referring to a concept by an attribute of it. For example, the crown referring to a monarch. See also synecdoche. |
| microcontext | A microcontext is a very small, specific context. |
| microcontext hypothesis | In linguistics (and many other things), representations of rules that are both detailed and broad tend to become very complex in order to characterize all of the exceptional conditions that can arise. As the context becomes smaller, the number of exceptional cases that need to be covered decreases. In sensemaking, the microcontext hypothesis says that representations (such linguistic rules for carrying out sensemaking operations) can be greatly simplified if they are required to work only in particular microcontexts. Simplicity arises for each set of rules because the number of exceptional cases that they need to handle is greatly reduced. In terms of research, the microcontext hypothesis suggests ways to divide up sensemaking problems into very small, manageable pieces. For example, term equivalence dictionaries may be amenable to broad use of statistical learning techniques if they use symbolic structures to create well-defined microcontexts. For another example, linguistic rules for building transformers may be kept very simple and use simple grammars (such as finite state grammars) if the field of application is delimited by appropriately-designed microcontexts. |
| multiple representations | Different representations have different advantages for displaying and manipulating information. For example, tables of attributes are useful for visualizing differences and outlines are useful for displaying narratives at different levels of detail. In the context of sensemaking, it is sometimes useful to have more than one representation for showing different aspects of the same information. Ideally in a sensemaking tool, the external representations are interlinked so that a sensemaker can switch back and forth between them and so that a change made in one visual representation of the information is automatically reflected in the other ones. |
| N |
| O |
| "oh no!" feeling | In the context of sensemaking, this is a feeling experienced when a potentially large or devastating difficulty is encountered in sensemaking. For example, an "oh no!" feeling may be experienced when substantial or awkward residue is found in the process of instantiating schemas ("residue tension"), or when a devasting counter-example is found for a theory or model. See also residue and "aha" feeling. |
| one-shot sensemaking | A sensemaking task that is brought to a single finish and then stopped. Contrasted with resumable sensemaking. |
| order | In the context of sensemaking, order refers to an organization or structure of information. For example, an order might be an organization of elements into disjoint or overlapping classes, arrangement into a taxonomy, an organization in terms of objects and attributes, organization in terms of a table, or a precedence relation. The term order is more general than the term ordering. See also structuring. |
| ordering | A precedence relation over members of a set, that is, a relation that indicates which members come before others. Mathematically, orderings are acyclic and usually transitive. Thus, if A comes before B and B comes before C, then (by transitivity) A comes before C. Orderings are often used in constructing outlines, narratives, and teaching sequences. For example, topics in a textbook are typically arranged in an order so that a reader will first cover the prerequisites for a topics before encountering the topic itself. See also partial ordering and complete ordering. |
| organizing | See structuring. |
| outliner | A program for creating an outline of a topic. An outliner could work by having a person work from partial outlines that are progressively detailed. It could also work by transforming some other representation (such as a graph representing a partial or complete ordering) and emitting an outline. |
| P |
| parser | A parser is a program that takes a set of sentences as input and identifies the structure of the sentences according to a given grammar. The term parser is sometimes used generically in cases where the sentences are made up of information units of any kind. |
| partial ordering | A partial ordering is an ordering relation that is not complete, that is, an ordering relation that fails to specify the precedence for some pairs, thus admitting more than one possible sequence for enumerating the set. For example, we might know for a given set that element A comes before J, and J comes before L, M, and N, and that M and N come before Z. In this example, we would not know which comes first between L, M, and N -- and it might not matter. A partial ordering can be represented by a DAG where arcs between nodes indicate which parts of the relation are known. |
| part/whole relation | A relation on objects which indicates when some objects are parts of others. For example, a door or window may be part of a room, and a room may be part of a building. In the context of information retrieval, part/whole relations can be used to broaden queries. For example, a query for documents discussing "government" decisions about telecommunications might be broadened to include "court" decisions about telecommunications on the basis that a court is part of a government. See also meronymy. |
| patch | In the context of information foraging theory, a "patch" is an area in an information space where there is a high density of relevant or high quality information on a topic. For example, a patch on the web may be a web page with many links to high quality information on a topic. On a file system, a patch may be a folder containing most of the relevant documents on a topic. Since there is no fixed notion of "space" in cyberspace, a useful way to characterize a patch is in terms of temporal distance rather than spatial distance. Thus a patch is a place where in a relatively short amount of time a sensemaker can encounter much useful information on a topic. |
| perceiving order | The operation of looking at information and noticing an opportunity for a new or different organization for it. This operation marks a transition in the perception of the information from chaos to order. For example, one might recognize that information could be organized in terms of disjoint categories, or in a taxonomy, or in terms of objects and attributes. Perceiving order is more general than parsing because it does not imply the use of a given grammar. In parsing, the grammar is given. In perceiving order, the grammar itself may need to be created or discovered. (Perhaps this creative act is related to the "aha" feeling.) Perceiving order is different from structuring, in that it refers simply to a shift in perception rather than a modification to an external representation. See also order and structuring. |
| phase shift | The term phase shift comes from the physical phenomena where a small change in a physical parameter causes a large-scale change in structure. For example, as the temperature goes below the freezing temperature, physical substances undergo a global shift from a liquid to a solid. The point at which the transition occurs is called the critical point. In the context of computation, a phase shift refers to a sudden qualitative change in the behavior of a process over a small change in one of its governing parameters. In the context of information retrieval, the term phase shift has been used to refer to a radical change in the number of matching documents retrieved over a small change in the query. |
| piles | In the course of office work, it is common for people to organize documents in terms of piles, such as an in-box, out-box, a pile of urgent items, and a pile of less urgent items. From a perspective of sensemaking, piles are an external representation with several interesting properties. At least in the physical world, documents can be in at most one pile at a time -- so that piles represent disjoint categories. Piles have a complete ordering. The top item of a pile is more visible than lower ones. Sometimes the edges of items in a pile can be seen. Items in a pile can be sorted according to some criterion. Virtual piles can also be created as a user interface. Virtual piles can be useful for putting documents or other things in disjoint categories. (They can also be extended beyond the properties of physical piles so that an object can be in more than one pile at a time, or can have corresponding "ghost" objects to represent it in more than one pile.) |
| polysemy | Refers to words that have more than one meaning. See also word sense. |
| precision | A measure of whether the documents returned in an information retrieval process are relevant to the query. For example, if eighty percent of the documents returned by a query are relevant (and the remaining 20 percent are not) then precision is .80. |
| Q |
| query-free retrieval | Query-free retrieval is an approach for retrieving documents in which the user does not need to make explicit queries. In systems that perform query-free retrieval, the system has some other means for determining what documents are potentially relevant, such as by examining the information context in which the user is working or by matching against other documents that the user is using. |
| R |
| recall | A measure of the percentage of the relevant documents returned by a query. For example, if the retrieval process finds only 60 percent of the relevant documents (and misses the other 40 percent) then recall is .60. |
| re-encoding | In a sensemaking task, effort is expended to represent information in a given form or schema. In the course of sensemaking, it often becomes useful to modify the schema. In this case, it is generally necessary to re-represent the previously encoded data in the new schema. Another way to say this is that re-encoding is the operation of transforming previously represented data to a new form or organization during a representational shift. The cost of such a transformation is called a re-encoding cost. See also representational shift. |
| refinement | In the context of sensemaking, refinement is a sensemaking operation that purifies and possibly grades information. The term is used in an analogy to the mining and refining of ore or oil. An oil refinery removes impurities from the crude oil. It can also split the oil into different grades -- which are used for different purposes. Information refining produces enriched information. See also document mining and sensemaking operations. |
| relevance feedback | A form of query-free retrieval where documents are retrieved according to a measure of equivalence to a given document. In essence, a user indicates to the retrieval system that it should retrieve "more documents like this one." |
| representation | A representation is a symbol structure that stands-for something. For example, the term "room temperature" may stand for the current temperature of a particular room. In sensemaking, the term representation is usually used to refer to symbols or objects in the workspace or document of the sensemaker. Thus, a sensemaker may use a table-based representation that stands for a set of objects and their attributes or an outline representation whose elements stand for topics in a report. Also see external cognition. |
| representational shift | The process of changing a representation in order to better accomodate the data that are encountered. For example, when there are relevant data encountered and no means for characterizing them adequately in the representation ("residue"), the schemas can be expanded. When data do not fit the established categories, the original schema or concept categories may need to be merged or split, or new categories may be added. |
| "rep shift" | Shorthand for representational shift. |
| residue | In a learning loop, residue refers to ill-fitting or missing data or unused portions of the representation. In the process of instantiating representations to capture data in comparable ways, residue refers to the data that arise that don't fit in adequately. Typically, the process of shifting a representation is guided by the discovery of residue. |
| resumable sensemaking | Resumable sensemaking is the sensemaking analog of life-long learning, that is, it embodies the idea that (at least potentially) the process of making sense is never done. New information may come in about the subject matter or even about the sources that were used in making the report. If the report is grounded in assumptions, new information may become available that contradicts or supports the sense. A resumable sensemaking system is one that keeps in place all of the links, term-equivalence dictionaries, and other information about the sensemaking process so that it can be readily restarted and reconsidered. Contrasted with one-shot sensemaking. |
| S |
| schema | A template for representing a class of data. Typically a schema provides a framework for naming and storing different elements of information about something. |
| segmenter | A program that identifies and extracts snippets from a document. Segmenters can identfy snippets using different kinds of knowledge, including not only punctuation (such as carriage returns), but also topical shifts in the word patterns. |
| semantic matching | A matching process for documents that uses knowledge of meaning to broaden recall. For example, semantic matching may use knowledge of synonyms, knowledge of part/whole relations, knowledge of class/subclass relationships, and knowledge of the user's information context to increase both recall and precision. |
| sensemaking | The process by which individuals (or organizations) create an understanding so that they can act in a principled and informed manner. Sensemaking tasks often involve searching for documents that are relevant for a purpose and then extracting and reformulating information so that it can be used. When a sensemaking task is difficult, sensemakers usually employ external representations to store the information for repeated manipulation and visualization. Sensemaking tasks inherently involve an embodiment as an actor (or actors), an environment, forms of knowing, and ways to work with what is known. Working can take different forms -- such as logical, metaphorical, physical, or image-based reasoning. See also external cognition. |
| sensemaking operations | A sensemaking operation is any of the information processing actions in sensemaking, whether it is performed by a person or a computer. A theory of cost structure in sensemaking is generally based on an analysis of the time required for performing sensemaking operations, a model of how the time varies with scale, and a model of the frequency of performing different operations in a certain kind of sensemaking. Different theories of sensemaking can have different sets of operations. Examples of sensemaking operations are abstracting, annotating, assumption linking, classifying, clustering, comparing elements or schemas, concept splitting, making a cross product, detailing, document mining, emitting, extracting, format stripping, foraging, fusing, goal shifting, instantiating schemas, linking, matching, negotiating meaning, perceiving order, re-encoding, refining, retrieving, segmenting, shifting representations, source linking, summarizing, stemming, structuring, transforming, and zoning. |
| sense map | A sense map is an external representation that maps snippets retrieved by document retrieval onto an evolving work space. In short, it indicates which snippets retrieved from the corpus are likely to be relevant to which parts of the workspace. |
| snippet | A segment of a document. Typically, a snippet is a set of contiguous text about the size of a paragraph and is about a single topic. Non-text items could qualify as snippets too, such as a graph, picture, or diagram. |
| source-linked sensemaking | In conventional media, a document about a topic may cite a list of references. In source-linked sense-making, the report is an active document with active links for retrieving the sources used. For example, it could include active links in each document element to the sources used as well as commentary about the sources. It also includes links to sources that were considered but ignored as well as annotations on those. Annotations on sources could include rankings and reasons for including or ignoring them. |
| source linking | It is common practice in writing many kinds of reports to indicate what sources were used for information. In the context of sensemaking, this is called source linking. If the sensemaking report is an active document and assumptions are represented explicitly, it is possible for a reader or sensemaker to go from elements of the report to supporting information sources. If a sensemaking system supports source-based reasoning, then sources may be marked as valid or ranked, and the display of the report can be adjusted to show which elements or conclusions are supported (or not) by a set of trusted sources, and which elements are supported only by untrusted sources. |
| stemmer | A program that identifies or extracts core roots from a word, removing prefixes and suffixes. For example, the words run, runs, ran, running, and runnable all have "run" as the root. A stemmer is often used in matching processes to make it possible to recognize that documents are about the same topic even when they use variants of the same words. |
| structuring | The operation of imposing an order or organization on a set of information. For example, structuring could be carried out by clustering or by extracting information from a document and putting it into a schema or by grouping documents into piles. |
| subclass relation | A subclass is a specialized version of a class. Consider, for example, that the class named animal represents the class of animals, mammal represents the class of mammals, and dog represents the class of dogs. In this example, dog is a subclass of mammal, and mammal is a subclass of animal. Conversely, animal is a superclass of mammal, and mammal is a superclass of dog. In the context of sensemaking, subclass relations are sometimes used to broaden queries. Continuing this example, a query to find articles about "pack behavior of mammals" might be broadened by including articles about "pack behavior of dogs." In AI literature, the name is-a is commonly used to refer to a subclass relation. For more technical jargon for this relation from linguistics, see also hypernym and hyponym. |
| subgraph | A connected subset of a graph. Typically, a subgraph is a connected subset of nodes and links from a larger graph. Graphs are a common representation in sensemaking where the nodes are frames and the links represent binary relations of some kind. |
| subgraph equivalence | In sensemaking, notions of equivalence are important for governing retrieval based on assumed similarity of meaning. Two subgraphs are equivalent if there is a mapping such that (1) for every node in the first subgraph there is a corresponding node in the second subgraph, (2) for every link in the first subgraph between two nodes there is a link of the same kind (or label) between the corresponding nodes of the second subgraph, and (3) for some meaning assigned to a node in the first subgraph, the same meaning is associated with the corresponding node in the second subgraph. |
| subgraph matching | The process of finding a subgraph in a larger graph, or finding a subgraph that is equivalent in a specified way to a given subgraph. |
| summarizing | Creating a short description of a document or information object that includes its most salient features for a purpose. Some computer programs that summarize text use statistical methods and document structure to identify and rank sentences in a document as candidates for inclusion in a summary. |
| superclass | A parent class of a concept. For example, mammal is a superclass of dog. See also hypernym and subclass relation. |
| synecdoche | Referring to a concept by a part of it. "All of the big names in the field were there." See also metonymy. |
| synonym | One of two or more words in a languge which have the same or very nearly the same meaning. In the context of sensemaking, it is important to note that whether two words are synonyms depends on context. For example, in many contexts the terms man and woman are not synonyms since they refer to opposite genders. On the other hand, in a gender-neutral context, both terms can be synonyms with each other and with person. See also synonymy, term-equivalence dictionary, and word sense. |
| synonym expansion | The process of expanding the set of terms in a query to include synonyms of the original terms. |
| synonymy | The synonym relation between words. Synonymy is not necessarily transitive in that a word can be a synonym of several other words that have very different meanings. For example, the word hot is a synonym of spicy, eager and warm. A human reader generally uses context to supply clues for disambiguating which meaning of a word is in force. See also polysemy |
| T |
| table | An organized external representation of information, typically in rows and columns. |
| tagger | In the context of linguistics technology, a tagger is a program that labels words in a document according to their role in sentences. For example, words may be classified according to their part of speech or other linguistic properties. |
| term-equivalence dictionaries | In information retrieval systems, queries are used to specify what documents are wanted. Typically, a computation is performed on the query terms to include equivalent terms. For example, words may be stemmed so that they will match against other forms. Beyond that, it is possible to use use other information to broaden the query, such as synonyms, part/whole relations, and class/subclass relations. Term equivalence dictionaries are dictionaries that give guidance on how to broaden queries. See also synonymy, part/whole relations and subclass relations. |
| transformer | In the context of sensemaking, a transformer is a program that converts information represented in one form to another form. A transformer program can be an important part of reducing re-encoding costs in sensemaking. |
| U |
| V |
| visualization | An external representation that makes it easy to see certain patterns in data. In the context of information systems, a visualizer is typically a computer program that presents data according to a particular pataern. |
| W |
| what-if sensemaking | See assumption-linked sensemaking. |
| word sense | A given word can have several meanings or senses. For example, the word hot can mean a high temperature, fiery, excited, eager, or pungent and spicy. A word sense is a given meaning of a word. See also polysemy |
| X |
| Y |
| Z |
| zoning | To narrow the selection of potential documents to be considered in a corpus. |