1 / 26

Automatic stylistic processing for classification and transformation of natural language text

Automatic stylistic processing for classification and transformation of natural language text. Foaad Khosmood Fall 2008. Overview – 3 big sections. Background Intro Applications Related work System Design Examples Tools Current Status and Future Plans Current status Evaluation Plan

zora
Télécharger la présentation

Automatic stylistic processing for classification and transformation of natural language text

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Automatic stylistic processing for classification and transformation of natural language text Foaad Khosmood Fall 2008

  2. Overview – 3 big sections • Background • Intro • Applications • Related work • System • Design • Examples • Tools • Current Status and Future Plans • Current status • Evaluation Plan • Work Plan

  3. Introduction/definitions • Style • Way of doing something (“option” – Wallpole) • Chess Opening • In Linguistics • Dialect • Register • Field (subject matter) • Tenor (formality) • Mode (medium, like “spoken”, “written”) • In other fields • Text type (in classification) • Helsinki corpus • Authorial style (in authorship attribution) • Stylometrics

  4. Detecting Style in Text • Using style markers • Features in text • Any recognizable component • Strictly defined, or rule-based / derived • Feature sets/combinations can model the lay notion of style • This project • Recognize, detect and measure markers • Add/remove markers through alterations to text without changing meaning.

  5. Applications • Digital Forensics • Style obfuscation and detection • Style-based search • Educational and productivity tools • Better writing assistants • Authorial tools, digital entertainment • Machine Translation (MT/SMT) • Interactive agents • For help, digital assistants

  6. Related work by others • Stylistics in Humanities • Literary studies (Bradford) • Controversy over meaning creation and usefulness (Simpson, Fish, Carter) • Style markers, computational stylistics • Wallpole (style as option, exercises) • Luyckx, Argemennon, Juola (stylometrics, authorship attribution) • Brill, Heidorn (writing assistants)

  7. Related work by us • Authorship attribution • Using neural nets (IJCNN 2005) • Unified methods and processes (ICMLC 06) • Transformation • NSF proposal (’06) • BCS paper (’08) on classification/transformation • WCECS (’08): System Engineering • DGfS (’09)

  8. System Objectives • Classification • Match text with corpus • Accept source, multiple targets • Transformation • Accept source, one target • Utilize library of markers and transfoms • Provide metric-based performance

  9. A trivial scenario in depth Overview • corpus, T, which can be described as modern, formal, factual writing with short sentences. • Marker detector functions: • M1() returns inverse of the average words per sentence (i.e. sentence/word). • M2() returns a ratio of overall instances of single digit numbers written using Arabic numeral symbols (i.e. “1”, “2”, etc.), to the total number of single digit numbers written in the text. • We are also given 2 transform functions: • X1() uses some symbolic linguistic techniques to split large sentences into two or more smaller ones. • X2() replaces instances of a number written in English to its equivalent in Arabic numeral symbols.

  10. Example - continued Finally, we are given the following text from a recent new story as source text (S): • Canada's opposition parties have reached a tentative deal to form a coalition that would replace Prime Minister Stephen Harper's minority Conservative government less than two months after its reelection, a senior negotiator said on Monday. This text contains 35 words and one of them is a number (“two”).

  11. Example -continued Analyze the target corpus, T (for this example we specify target corpus characteristics) • M1(T) = 0.2 • M2(T) = 0.8 • Analyze the source text, S • M1(S) = (1/35) = 0.0286 • M2(S) = (0/1) = 0.0 • For comparison and evaluation, a simple root mean squared error (RMSE) is calculated between the S and T vectors. • RMSE(S,T) = √(( 0.2 - 0.0286)2 + (0.8 - 0)2) = 0.8182 • The system determines the RMSE distance is too high to constitute a match

  12. Example -continued • Operator X1 is applied to source text, S • Operator X1(S) converts the text to version S2 (inserted words and punctuation are underlined) • Canada's opposition parties have reached a tentative deal. The deal is to form a coalition. The coalition would replace Prime Minister Stephen Harper's minority Conservative government. This is less than two months after its reelection. A senior negotiator said it on Monday. • S2 is analyzed • M1(S2) = ((1/8)+(1/7)+(1/11)+(1/9)+(1/7))/5 = 0.1225 • M2(S2) = (0/1) = 0.0 • New RMSE is calculated • RMSE(S2,T) = √(( 0.2 - 0.1225)2 + (0.8 - 0)2) = 0.8037 • System determines RMSE is still too high so we continue.

  13. Example - continued • Operator X2 is applies to source text, S • Operator X2(S2) converts the single instance of “two” to “2” and produces S3. • Canada's opposition parties have reached a tentative deal. The deal is to form a coalition. The coalition would replace Prime Minister Stephen Harper's minority Conservative government. This is less than 2 months after its reelection. A senior negotiator said it on Monday. • S3 is analyzed • M1(S3) = ((1/8)+(1/7)+(1/11)+(1/9)+(1/7))/5 = 0.1225 • M2(S3) = (1/1) = 1.0 • New RMSE is calculated • RMSE(S3,T) = √(( 0.2 - 0.1225)2 + (0.8 - 1.0)2) = 0.2144 • At this point the system can either determine that this RMSE number is within tolerance levels of the target corpus, or it can equally decide that since the transform functions have been exhausted S3 represents a best effort transformation for this problem.

  14. Example - conclusion • Weight distributions among markers • Operators applied successively • Optimization is NP complete

  15. System design

  16. Components • Analyze source • Analyze target • Apply transforms • Classify • Compare • Evaluator • Operator library • Target corpus • Transformed text

  17. Operators • Two example types • Surface replacement • May require part-of-speech tagging • May require dictionary selection • Acronyms • Semantically aware • Syntactic parsing • May require scanning of whole document • Yoda

  18. Resources and tools • AAAC / JGAAP • Patrick Juola’s Duquesne contest (AAAC) • LinkGrammar Parser • Paraphrasing / summarization tools • WordNet / Ontology / Word lists • Synonyms and related words • GNU diction, style, flex, bison • Readability measures • Lancaster University course on Stylistics

  19. Current Status • A series of scripts • getMarkers(Text) • analyze(Text1, Text2) • transform(Text) • Missing capabilities • intelligent selection of markers and transforms • stager or planner of operations • modular, web accessible contribution • optimization for speed

  20. Another example with current tools

  21. Target – Elizabeth I speech in 1588 Elizabeth I of England - 1588

  22. Results • Original source: • I am $AI4, of the $CIVADJ7. We give you a choice: peace and friendship - or blood, sweat, toil and tears. Please accept the following exchange as a sign of your friendly intentions toward our people. • Transformed source: • I am $AI4, of the $CIVADJ7; we, under God, give you a choice: peace and friendship , or blood, sweat, toil and tears. Take this exchange as a sign of your friendly intentions toward our loving people.

  23. Classification-Transformation Loop Method that can serve as a process directive • Classification and Transformation are intimately related • Classification as way to detect Transformation • Inadequate transformation => richer markers • Richer style markers => more precise classification CLAS TRAN

  24. Evaluation • Classification • Done by classic supervised machine learning techniques • Multi-fold, labeled corpus • Transformation • Machine verification • Auto classifier of same system • External classifier • Human verification • Periodic (perhaps 2 per phase) human assessment of classification

  25. Timeline • Phase I: Vertical prototype (3-5 months) • Build the skeleton of the prototype system using the new programming tools and web environment. The goal is to get it to the level of the scripting based system we currently have. Very few markers and very few transformers will be implemented other than some token ones to demonstrate the feasibility and modularity. • Install and facilitate database communication with the web based system. • Gather appropriate corpora for testing. • Phase II: Build up collection of markers and transforms (3-5 months) • Include more markers and transformers using researched resources and techniques. Consult more linguistic experts and tools in order to develop more robust and more generic natural language manipulation tools. Markers should be empirically tested in strict classification exercise to determine their relative values. • Work on planning algorithm for transformer selection, as well as statistically based markers (i.e marker functions that are based on specific text features and will be different for different corpora). • Phase III: Evaluator, comparator and planner (3-5 months) • Work on automatic normalization and weight adjustment for the marker vector based on reinforcement learning. • Work on alternative and interchangeable distance metrics between texts (as opposed to strict style matrix RMSE). • Work on abstracting the planning function and allowing user control over its operations. • Phase IV: Transform granularity and constraint satisfaction (3-5 months) • Develop a granularity operator for transforms or degrees of freedom (i.e. even though applicable to many sections of a text, the transform functions should manipulate very few sections of the text at one time) and a measure of granularity should be specified and utilized by the system. • Improve transformation performance by adding more constraints for basic meaning transference (i.e. grammatical correctness and measurable meaning orthodoxy). • Test transformation with more corpora and validate by using best classification algorithms. • Develop workable prototypes for one or more real-world applications.

  26. Questions, more demos

More Related