Category: work

David Hirtle is doing graduate work here at the University of New Brunswick, in computer science. The full title of his paper is “TRANSLATOR: a TRANSlator from LAnguage TO Rules.”

Semantic web is still n ot widely used.

– Focus of current development: machine-readable (meta)data

– Problem: only experts can contribute. Need to lower barrier to entry.

Provide a user-friendly format!

– why not English [he really means natural language]?

– “controlled English” avoids ambiguity: it’s formal, but also natural

TRANSLATOR will translate “every student gets a discount of 15 percent” to express [in XML, from what I see] that “student” implies “customer,” etc.

ACE (attempto controlled english):

– looks like English: “every honest student who does not procrastinate receives a good mark and easily passes the course.”

– but actually a formal language, like RDF: a tractable of English – all ACE sentences are English, but not vice versa

– every ACE sentence can be unambiguously translated into logic.

Strategies for handling ambiguity:

– exclude imprecise phrasings (“students hate annoying professors” – do they hate to annoy profs, or do they hate profs who are annoying?)

– interpretation rules (“the student brings a friend who is an alumnus and receives a discount” – who receives the discount? in ace, by default, it’s the student because of a certain rule. If you want it to be the alumnus, you write “…and who receives a discount.”)

How can rules be expressed?

– in natural language, many different forms (everyone is mortal, all humanity is mortal, for each person the person is mortal)

– all above are valid ACE

– further embellishment (negation, relative clauses, etc) [vz: but doesn’t that add ambiguity?]

What can’t yet be easily expressed?

– “infix” implication (“the student is happy if there is no class” – solution: TRANSLATOR swaps the condition(s) and conclusion(s) and voila, ACE-acceptable)

– production and reaction rules (involve actions: “if a student is caught cheating then send a report to the registrar” requires the imperative mood, which is not yet in ACE)

Discourse representation structures, and more technical info. Sad, I can’t reproduce his diagrams here. The rules are eventually translated into RuleML, in whose development David is participating.

RuleML:

– goal is interoperable rule markup (XSLT translators to other semantic web languages)

– family of “sublanguages” (modular XML schemas; each represents a well-known rule system; TRANSLATOR uses First-Order Logic sublanguage)

Why use RuleML?

– ease of interchange (XML)

– compatibility with RDF and other languages, as well as W3C’s upcoming Rule Interchange Format

– availability of tools

– wide fariety of features (negation-as-failure, weightings, data types etc.)

Again, work-in-progress. Truly an attempt at getting closer to the semantic web. Formalizing natural language, what a gargantuan task. One critical benefit of TRANSLATOR is that it “allows non-experts to write facts and rules for the semantic web.” When can we play with it?

Now, it appears. Here’s a site for TRANSLATOR, including a Java Web Start demo.

Shelly Lukon and Patrick Juola are both at Duquesne University. The full title of their paper (presented by Lukon) is “Designing a context-sensitive machine-aided index generator.”

Problem definition.

Back-of-the-book indexing provides relevant terms, identifies cross-references and subcategories, and has a static, rigid structure (as opposed to web indexing). Human indexers invest a LOT of time into indexing (1 week per 100 pages of text); use software to automate mundane texts; and make all the intelligent indexing decisions. SL&PJ’s prototype system bridges the gap between the human and currently available tools, but not to replace the human indexers.

They’ve interviewed professional indexers, product-tested some of the software packages they tend to use, and looked at some mathematical techniques (particularly LSA, latent semantic analysis) that have had proven success in text processing and capturing semantic content of terms.

Cognitive tasks involved in index construction:

– identifying terms to index;

– locate all informative references;

– identify/locate synonymous terms;

– split index terms into subterms;

– develop cross-references within text;

– compile page numbers.

Their techniques for obtaining semantic information:

– parsing/tagging of terms, frequency analysis

– LSA

– word sense disambiguation (WSD)

– hierarchical cluster analysis (HCA)

This is still a work in progress. So far they’ve been able to locate all informative terms in text, and to allow the user to set thresholds/parameters. LSA, WSD and HCA show first level of clustering nearly 40% accurate upon inspection (not great, but a solid start). Their single-processor PC takes several hours to process small (60K words) corpora. Better than a human’s speed!

They’re categorizing words into parts of speech: identify the part-of-speech of each term; label each term with delimiter and acronym (home becomes home/NN since home is a noun). They’re only dealing with English right now. Their app is written in Java, as is MontyLingua which they’re using for part-of-speech tagging.

LSA:

-use factor analysis to generate numerical representations of terms and their meanings;

-divide corpus into “documents” (paragraphs), then analyze each unique “term” (word) relative to each document;

-create term-by-document matrix;

– create term-by-term covariance matrix (look at how each pair of terms vary together)

– singular value decomposition (SVD) – a way of explaining variability among random variables (dimensions)

– decompose covariance matrix into three submatrices [over my head here]

– rank resulting values

– reconstruct using most significant dimensions (reduce noise, sharpen similarities/contrasts)

– 200 most significant dimensions: pinpoint each term’s location in 200-dimension “semantic space” [why 200?]

WSA

– separate out different senses (meanings) of each term token

– numerical encodings generated by LSA give average context for each term token

– look at encodings of the other terms surrounding each occurrence of the token

– Example: the word “bass” occurs throughout text (both as fish and as musical instrument), proximate to other words (guitar, boat, fish) that help disambiguate

– disambiguate “bass” into “bass_fish” and “bass_instrument”

HCA

– partition terms into subsets with similar properties/characteristics

– antonyms as well as synonyms will cluster together (both have strong relationships, but the system doesn’t know whether they’re positive or negative)

– this information can be used to identify cross-refs (see also) and subterms

This is a machine-aided system. Its purpose is not to replace but to assist the human indexer, whose judgment and experience cannot be fully captured by a sophisticated expert system. Users can edit results at any stage, control indexing parameters, etc.

Metrics for evaluating the “goodness” of the resulting index:

– side-by-side comparison between entirely-human-generated and machine-aided indexes of the same dataset, quantify what percentage of agreement is acceptable, maybe find meaningful information in how they disagree as well.

Future work:

– incremental refinement

– system has modular architecture for ease of swapping out individual components

– need robust, effective user interface

– empirically vary frequency thresholds, weighting methods, number/percentage of dimensions to use in the reduced data matrix

– continue to build in the latest/most efficient indexing/retrieval methods.

What a great project. I’d love to use it for RolandHT, but it probably won’t be done in time. Enabling the software to read/process XML is on their wish list of big enhancements, hooray!

[Oh look: Geoffrey Rockwell is posting some of his thoughts about this CaSTA conference on the TADA wiki. Highly recommended reading.]

Ian Munro is the Canada Research Chair in Algorithm Design at the Univ. of Waterloo. The full title of his keynote is… well, in the schedule it’s “Computer science research for text analysis,” but on the opening slide it’s “Developing text analysis software.”

Will talk about text search, one of his interests. He’s hard-core CS.

The need for computing in the humanities: “Scholarship increasingly depends on electronic document repositories and the growth of digital libraries… Even more apparent in computer readable form are collections of business documents and linguistic corpora. Gray literature, including technical reports, personal communications, and online help information, also constitute a growing text source.” –Frank Tompa

IM’s resaerch: data structures. How to organize information so we can find what we want: quickly; using an acceptable amount of space; proving the necessary inherent time and space bounds. [vz: bless his heart.] He’s on the theoretical side of computer science, a very different side from “user interface” or “understanding natural language” sides.

Where di IM get going on text? The Oxford English Dictionary project; interaction with humanists and lexicographers. New problems to work on; great data.

Another project IM was involved in, in the early 1980s: Videotext. Like the internet, but assumed few information providers, and access would’ve probably been restricted. The software ideas were there, but it was too early to use them.

IM gives some history of the OED project, which is actually covered pretty well in the Wikipedia article about it. The article includes a description of the first SGML encoding(s) of the OED.

The software they developed for the OED project:

– Lector, a general purpose browser. Worked with tagged text, presented in reasonable form, early SGML that, were they doing this a bit later, would’ve been HTML.

– Goedel, a programming language/database system.

– Pat, a search engine.

Pat is short for PATRICIA, “Practivel Algorithm to Retrieve Information Coded in Alphanumeric.” [vz: oy!] It does full-text searching, using an approach now generally known as “suffix tree.” In fact, in the final implementation it was a “suffix array.”

Typical problem: text indexing. Let’s take a large text file, like all the documents/email for a company, or a genome. We need to construct a structure so that given an arbitrary phrase they can quickly find where this phrase occurs in the “document.” Call the “extra stuff” an index.

What’s “suffix array”? It’s a method: an array of pointers referring to text positions in lexicographic order. Allows binary research. More on it here. (By the way, about this and other links: yeah, it’s wikipedia. Don’t even start with me on it being a Bad Resource. It’s not, unless you take it for gods’ word.)

Then IM describes suffix tries. This is all so far over my head that I’m not even going to try to summarize it; besides, the link does it pretty well.

From the OED project, IM and colleagues’ work proceeded to:

– more text search;

– data warehousing for asking complex queries

– enabling people to view relational databases as text (tags substituted for fields)

– enabling people to get things in “sorted” order: online phone books; buildings wired separately [tell me the companies that have offices in buildings I, a phone company, have wired – but who are not yet customers of mine); Sarah Lee = Sara Li; Romeo and Juliet (how many places are there in England where someone named Romeo lives near someone named Juliet? IM says that the answer is three.)

Where do things go next? IM wants to get rid of the tedium of searching in raw form, scanning texts etc, all parts of humanities work; improve the language interface; utilize better OCR (optical character recognition); build an application that can handle archaic linguistic forms.

Missed the very beginning. Stan Ruecker again; Milena Radzikowska (Mount Royal College); and Stéfan Sinclair (McMaster University). “Communicating process with form: designing the visual morphology of the Nora data mining kernels.” Stan presents the new Nora interface. Which is VERY pretty, but sadly not viewable online (but see the Nora link below).

Rich-prospect interfaces:

– some meaningful representation of every item in the collection

– tools for manipulating the display

– tools shiould rely on information emergent from the collection

– [missed the other three]

Introduces NORA.

Purpose:

– classification and pattern recognition (“find me documents that I’ve identified, and tell me the features that you’re working with.”)

– “specifically, we aim at allowing literary scholars to identify categories of interest in collections of literary texts, and then to find new members of those categories and explore the features that correlate with those categories…” -John Unsworth

“Kernels” are objects that you’re working with. As you group cultural artifacts together and define some criteria for grouping, the kernel’s (initially blank-outline) iconocragphic representation changes, becomes more complex. Each kernel has five different states, from blank to very complex: leaves, snowflakes, concentric circles and other metaphors – they don’t seem to have settled on a single metaphor yet, but they’ve agreed that five states are definitely not enough.

You generally have multiple kernels, with a pallette for each; so you can see a bird’s-eye view of all your kernels. You can also “save” a kernel by dragging it to the “deasktop” environment (this is all happening in your browser), and then you can share that kernel image with others or refer to it in your own research later.

Sort of a Photoshop for text analysts. With much, much easier controls. I’m looking forward to more development of this!

NORA (No One Remembers Acronyms) is merging soon with MONK (Metadata Opens New Knowledge). Soon they’ll all be known as MONKeys!

Last paper of the day – the rest will be posters. see y’all tomorrow.

Terry Butler is at the University of Alberta. The full title of his paper is “Automated indexing using an existing thesaurus: a bridge between Coleridge and Roget.”

The project aims to usefully present a digital edition of Samuel Coleridge’s notebooks. They’ve been printed, with annotations, and published by Princeton. The notebooks are published with name and place indexes, but without a subject index (a big drawback for researchers). Current project is focusing on building an electronic subject index to the print edition.

The complexity of transforming handwritten notebooks into a printed format is staggering. Butler et al. are digitizing the printed edition, and semantically encoding it. The encoding captures not only the structure of the text, but also its dynamic nature – the deletions, insertions etc. so strongly present in Coleridge. Plus a thematic index. They’re also marking up foreign-language material as being in other languages.

Roget’s thesaurus was first published in 1852, and he’d been working on it for 30 years. So the thesaurus is roughly contemporary with Coleridge’s writings, and word similarities imply similarities in thinking and interests between the two men. They want to link up Roget’s broader categories to instances of writing in Coleridge, as a starting point to studying the notebooks.

They take all of the words from each of the notebook entries, stripping out metadata and foreign words, and now they have a “bag of words.” They then stem the words to get more matches between them and the thesaurus. They want to find out connections: which of these words also occur in Roget’s Thesaurus (25K+ words, 1K+ headings, organized into 6 large classes)? Results are displayed in both directions: all the Roget connections to a given note; and all the notes connected to a particular Roget entry.

Value of automatic linking to scholarship:

– provides access for searching that is complementary wto what the text itself says

– text is related to larger, consistent conceptual categories

– it can be developed (from the electronic text) in a few hours’ [of computer processing] time: a custom thesaurus would take hundreds of hours to create.

Limits of Roget:

– it’s a verbal construct: not really about the world, but about words

– it has a lack of proper nouns

– it contains a large quantity of rare, obscure, dated words.

Assessing effectiveness:

– a subset of the results was given to expert team members for review: will this be a useful first step to creating a Coleridge thesaurus?

– links are distributed across information space [vz: ?]

Next steps:

– develop a [complete] customized Coleridge thesaurus;

– attempt to link between Roget and another contemporary text.

Renita Machado and Robert Murimi are in the Dept. of Electrical and Computer Engineering at the NJ Institute of Technology in Newark. The title of their paper, presented by Machado, is “RealBook – Reading a book through the eyes of reality.”

RealBook focuses on the use of wearable computers to convert the experience of reading a book into images and sound. They propose the use of an augmented reality system made up of a head-mounted display [with camera sensor], headphones, a physical book, and RealBook software. The latter is meant to augment the pleasures of reading a book.

They haven’t built it yet, but the most interesting thing they want it to do is convert events in a book images and sounds. They’ll use animation, which they test by performing participatory design (preliminary studies that produce, effectively, storyboards). The software needed will need to display enough animation to convey the meaning/flow of the book.

They want to encourage healthy competition among potential manufacturers of this thing. It could potentially apply not only to fiction but to non-fiction as well – textbooks, for example! Interesting.

The closest resemblance to this work is Magic Book, a tool to view illustrations in books in both augmented reality and virtual reality views. RealBook would be primarily entertainment, though, Machado emphasized.

[vz: sounds fascinating. Sci-fi. I want to play with this thing, and am impatient for it to get built.]

Willard McCarty is “a scholar-provocateur,” in the words of David Gants, who is introducing him at the moment. The full title of his keynote is “Beyond the word: Modeling literary context.”

WM is “the heavy metal version of an intellectual butterfly.” That’s in his words. Same disclaimer as with Shillingsburg: dense, dense. Choppy synopsis.

The humanist perspective on tools problematizes them, which is the point of tool development in the humanities. Our job is not to solve or to fix, but to question. Context explains nothing: it is merely more text. And yet, it remains: how else does the textual environment select for meaning, beyond contiguity and syntax?

Modeling defines an intellectual space within which we can apply styles of scientific reasoning without compromising the humanities.

Main kinds of text analysis, and their origins in prior habits of mind:

– Corpus linguistics (main foci: corpus, concordance, collocation; “words are known by the company they keep” [quote from Firth?])

– concordance is more than “known by the company it keeps”. its 17th-century practitioners were thinking typologically.

– Statistics: searching for hidden patterns of meaning

– some working patterns are inherently stochastic, essentially guesswork. (Humanities are intimately tied to stochastic ways of working, which isn’t to imply that sort of work is easy.)

– Metatextual [semantic] encoding: nontextual entity X exists in textual location Y. Gain control of the world by categorizing its parts.

– Relational database design. Resolves an artifact of study into one or more entities sharing an identical set of attributes. Strict logical principles apply, but there’s a trade-off between these principles’ limiting nature and the manipulability they allow for over significant amounts of data.

What does this analysis seek to do? [Here follows a complex argument which, when the webcast of the keynote goes up, I hope to fill in.]

Not building digital replicas of books (a literary jukebox), but modeling (purposeful gerund: the process of modeling!) of cultural artifacts is the point of humanities computing. Computers change the speed at which modeling can happen, and the amount of data that can be modeled. We think differently when more comes more quickly into view. A gap separates, and will always separate, modeling from the model.

The assumption that individual agents of a model are independent variables that don’t affect each other but act separately upon a given entity is erroneous and damaging to research.

Modeling makes our readings more informed.

…As usual when listening to Willard speak, I’m too overwhelmed by all the information flying at me. Consider this a placeholder.

Stan Ruecker, Sara Dorow, Heather Jiang, Zachary Devereaux and Christopher Moore co-wrote this one. The full title of their paper is “Shallow, narrow, deep and wide: issue crawling for Chinese adoption.” [vz: a bit confused as to what this means. Intriguing, though.] All authors are from the University of Alberta, with the exception of Zachary Devereaux who is from Ryerson/York University. The paper was presented by him and Christopher Moore.

ZD shows us hyperlink analysis maps, which he proposes are textual maps.

IssueCrawler: server-side webtool, created by the Govcom.org Foundation in Amsterdam. Crawling, as in spider robots that take as starting points the websites you’re interested in exploring. The example issue is adoption of Chinese children by people all over the world, particularly in the US and Canada.

To launch the crawl, the researcher sets the number of iterations of method (how far the crawler goes – how many websites) and depth (how deeply within each site the crawler explores). We go on to a demo of the adoption crawl (a shallow one – two sites out, two levels deep). Many .gov/.ca sites with factual information, and .com/.org sites with much more personal accounts (“pink and fuzzy,” they say, especially since 95% of Chinese children who get adopted are girls), acculturation tips, etc. Not very many Chinese sites in there, although there are some in the larger network. Third crawl is deep and wide: 2 iterations out, 3 levels deep, with legend: considered by research team as the most representative network map. Shows a “strong” inner network; presence of professional, resource, and secondary activities for adoptive families (consumer/charity/cultural). Highlighted is the Association for Research in International Adoption.

The researchers’ chosen topic highlights that it’s possible to hyperlink-analyze from a censored and uncensored point of departure, see the differences easily.

Conclusions:

– websites in this context represent dynamic and interactive hypertext.

– the result is an interactive and informative set of dynamic maps that aid analysis of the online presence of the communities in question.

– future research: cross-platform content analysis of network, as well as its different instantiations over time.

[my talk went pretty well. whee! freedom for the rest of the conf! love when I’m up early.]

Jane Morris is at the University of Toronto. The full title of her paper is “Readers subjective perceptions of lexical cohesion and its implications for computers’ interpretations of text meaning.”

[vz: interesting. I’m not sure what I think about the phrase computers’ interpretations. slippery slope from here to computer sentience, which I certainly ain’t against, but which is a controversial topic at best.]

Meaning of text can be approached from three different (and much debated) points of view: what the readers think it means (attentional structure), what we think the author(s) thought it meant (intentional structure), and what the text itself means (linguistic structure).

Machine text and corpus analysis (computational linguistics) can only access the meaning inherent [?] in the text, but not the readers’ or writers’ perspectives. Computational linguistics has often proceeded as though the attentional and intentional structures of texts didn’t exist.

Properties of text, according to David Olson: it’s an artifact; it’s a representation of knowledge/meaning; and it needs to be interpreted by readers.

Morris’ research question: how subjective is the interpretation of lexical cohesion of text? Investigated this in a study using 26 readers and 3 texts.

Some definitions. Lexical cohesion: contribution to a text’s meaning by groups of related words running through it. Linguistics studies: meaning, coherence. Computational linguistics studies: structure of text, summarization, information retrieval, spelling correction.

Study’s objectives: summarize agreement between readers on work groups, using individual diffs as an indicator of objectivity. Readers underlined groups of words they thought were related in the same color, different groups being different colors. They then wrote out those word groups separately, and wrote a brief description of what they thought the groups meant. (Funeral, communion, chapel, deceased: processes involved when someone dies.)

Study showed about 40% individual difference in readers’ lexical interpretation of texts. This implies significant limitations in machine analysis of texts. How to computationally account for differences interpretation? Morris proposes developing reader and writer models [vz: agent-based modelling might be useful?] Simple reader model: reader-specific corpora, thesauri, and view of lexical cohesion (use reader-based thesaurus – which can be generated using off-the-shelf software – to identify word groups in text).

Future research: different texts and different readers; subjectivity of other aspects of meaning; what does subjectivity mean or reflect (reader attitudes); [more on] how to create reader models, and also writer models to reflect writer subjectivity.

Peter Shillingsburg teaches and researches at De Montfort University in England, formerly of the University of North Texas.

Whoo, this is dense, and he’s reading – not much buffer talk. My synopsis is going to be incomplete and choppy.

Claims to be neither a computer scientist nor a humanities computing expert (the bulk of the attendees at this Breadth of Text conference.) He is one of the most prominent text analysts of the current scholastic age. “My field represents the problem, and your fields represent the solutions,” he says.

Question being addressed: what is it about what he does [we do] professionally that actually matters? Higher education isn’t aiming to create better citizens these days, so much as serving as “tertiary education” after which one learns how to live. Given the present [rather elitist] state of higher education, what is it that textual critics do that really matters?

It seems that preparation for career is at least more pressing, if not more important, than preparation for life. We work at a time when scepticism about what the university does is high, except insofar as it teaches money-making.

Teaching is said to be the second oldest profession in the world, at its height in the now-destroyed Library of Alexandria. Because texts are multiplied for dissemination, wear out and need to be replaced, copying texts is important – and highly imperfect (see also parlor games of telephone). Proofreading, then, is the first function of textual criticism, and otherwise excellent scholars seem to be terrible at it (editors, binders etc. of scholarly editions not excepted).

Then there’s the task of restoring integrity to texts that have fallen afoul of their original versions. Here [he implies] we need collaboration – among editors, philologists, historians. Shillingsburg proceeds to give us some rather egregious examples: one is the relationship between two Victorian-novel protagonists. Early in the novel, between 1850s (?) and 1989, they go up some stairs “hand in hand,” which was thought to foretell the development of their relationship. But the manuscript actually says that Henry went up the stairs hat in hand. This sorta changes the entire relationship, as portrayed.

What does one do when a text, as it exists, doesn’t seem to make any sense – but for which no alternative (manuscript, for example) exists? Some editors have deleted phrases or changed words, based on their own common-sense judgment. Who’s to do better? And what about editing texts by genius authors who use[d] words we have not yet learned (sprinkt for sprinkled)? There was a time when such things were “corrected” by well-meaning but erroneous folk.

Is it okay for a university to set up shop and do the best it can with what it has? Without a hard “standard” for editing ambiguous texts? [vz: this standard, even if it exists, is often meaningless!] What does “integrity” of a text mean?

The task of the textual critic, and importance of textual criticism, is [in] validating and maintenance task preliminary to actual criticism. Textual integrity is a foundational myth. There are indeed wrong ways to edit texts; but is there only one right way to edit them? The questions should be why the text has acquired its particular forms, who has made it so and under what cultural circumstances? [vz: I’m paraphrasing like crazy here.]

Textual criticism insists that any text will do – but that every text will do only that which it is capable of doing. It doesn’t necessarily represent “what the author thought,” for example. The task of the textual critic is to help students and teachers to know which text of a work they’re using, and to show how multiple/unstable/created texts change our perceptions of the work. A textual critic is proofreader, amender, identifier of the salient features of a work, who lets readers know what particular text of the work they have in hand, and what its particular characteristics are.

“Of him that is appointed to teach, the first business is to learn, an unintermitted attendance to reading must qualify him to be heard with profit….For such service he can be fitted only by laborious study, and study therefore is the business of his life; the business which he cannot neglect without breaking a virtual contact with the community. Ignorance in other men may be censured as idleness, in an academick it must be abhored as treachery.” –Samuel Johnson

Let’s not allow Johnson’s stirring words to mislead us regarding what academics can/cannot, should/should not do. What we do is, at the end of the day, criticism and not fact; inference and not truth. It is because of this inherent uncertainty, however, that our commitment to accuracy and precision should be greater. Ignorance is the academic’s most intimate familiar, never to be banished – only used to learn more about how much we have yet to do.