Differences
This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision Next revision Both sides next revision | ||
user:hladka:playcoref [2009/02/25 12:19] hladka |
user:hladka:playcoref [2009/03/11 15:00] hladka |
||
---|---|---|---|
Line 6: | Line 6: | ||
Určitě chceme mít jakékoli anotace, pro které se rozhodneme (tedy i koreferenční), | Určitě chceme mít jakékoli anotace, pro které se rozhodneme (tedy i koreferenční), | ||
- | |||
- | ====== Specification ====== | ||
- | |||
- | |||
- | |||
- | |||
- | |||
- | |||
- | |||
- | |||
- | ===== Strategy ===== | ||
- | Hook up the words which refer to the same object (entity). | ||
- | |||
- | A game of two players. | ||
- | |||
- | Session time up to 5 (???) minutes. | ||
- | |||
- | At the beginning, two first sentences of the document are displayed to each player. The players link nouns and pronouns which refer to the same object (an example TBA) independently of each other. When a player hooks up all the related words (s)he asks for a next sentence of the document. | ||
- | |||
- | |||
- | |||
- | ===== Input Texts ===== | ||
- | |||
- | === Text Selection === | ||
- | * Anja's data ## // PDT data that are currently being annotated for the extended coreference // | ||
- | * more ' | ||
- | |||
- | === Coding === | ||
- | * utf-8 | ||
- | |||
- | === Internal format === | ||
- | * sgml ## //propose dtd file: include the element '' | ||
- | * conversion: csts <-> pml m_coref scheme | ||
- | |||
- | === (Pre)processing === | ||
- | * tagging | ||
- | * t-parser by Linh | ||
- | |||
- | === Text handling === | ||
- | * sentence by sentence | ||
- | * supervised selection of documents for a session | ||
- | |||
- | ===== Scoring ===== | ||
- | * top lists | ||
- | |||
- | |||
- | ===== Output Data ===== | ||
- | |||
- | ===== Design ===== | ||
====== Motivační publikace ====== | ====== Motivační publikace ====== | ||
Line 64: | Line 15: | ||
* Luis von Ahn, Laura Dabish: //Labelling Images with a Computer Game//, 2004, ([[http:// | * Luis von Ahn, Laura Dabish: //Labelling Images with a Computer Game//, 2004, ([[http:// | ||
- | ===== Anotování koreference v českých datech ===== | + | |
+ | ====== Anotování koreference v českých datech | ||
* PDT 2.0 [[http:// | * PDT 2.0 [[http:// | ||
* rozšířená koreference - viz přehled [[https:// | * rozšířená koreference - viz přehled [[https:// | ||
Line 71: | Line 23: | ||
- | ===== Automatické určování koreference v českých datech - přehled ===== | ||
- | * Dosavadní experimenty | ||
Line 78: | Line 28: | ||
+ | ====== Automatické určování koreference v českých datech - přehled ====== | ||
+ | * Experiments with Czech so far | ||
+ | - Nguy Giang Linh: Návrh souboru pravidel pro analýzu anafor v českém jazyce (A set of rules for anaphora resolution in Czech), MFF UK 2006. **Available: | ||
+ | - Nguy Giang Linh; Žabokrtský, | ||
+ | * Linh's procedure | ||
- | ===== Návrh hry - brainstorming ===== | + | |
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ====== Návrh hry - brainstorming ====== | ||
**26/5/08 Anja, Bára:** | **26/5/08 Anja, Bára:** | ||
* Vstup: Texty v povrchové podobě, tedy NE tektogramatické stromy | * Vstup: Texty v povrchové podobě, tedy NE tektogramatické stromy | ||
Line 108: | Line 72: | ||
* **Vystup:** dvojice, na kterych se hraci shodli a neshodli. Shoda bude prinosna, ale co delat s temi dvojicemi, ktere oznacil pouze jeden hrac? Muze se objevit jeste jedna sance na shodu, a to ta, ktera prijde z jine partie se stejnym dokumentem. Proto bude muset byt kladen duraz na rizene podsouvani dokumentu do partii. | * **Vystup:** dvojice, na kterych se hraci shodli a neshodli. Shoda bude prinosna, ale co delat s temi dvojicemi, ktere oznacil pouze jeden hrac? Muze se objevit jeste jedna sance na shodu, a to ta, ktera prijde z jine partie se stejnym dokumentem. Proto bude muset byt kladen duraz na rizene podsouvani dokumentu do partii. | ||
* Kontrolni mechanismy, aby hraci nespojovali libovolna dve slova - vzhledem k navrhu na pocitani skore, tj. za shodu, budou hraci aspon castecne motivovani hrat smysluplne. | * Kontrolni mechanismy, aby hraci nespojovali libovolna dve slova - vzhledem k navrhu na pocitani skore, tj. za shodu, budou hraci aspon castecne motivovani hrat smysluplne. | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ====== Specification ====== | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ===== Strategy ===== | ||
+ | * **Hook up the words which refer to the same entity.** | ||
+ | * A game of two players. Players are paired randomly. Computer as a player: automatic coreference resolution **???????** | ||
+ | * Session time up to **???????** minutes. | ||
+ | * At the beginning of the game, if there is no coreference pair in the first two sentences (as determined by the manual/ | ||
+ | * What my partner is doing? If (s)he hooks up the same pair of words as I hooked up then the pair of words starts **??????? | ||
+ | * The players can re-hook up any word any time in the session. | ||
+ | * To design the game for a particular language the following data and tools are needed (or, better said, are welcome): | ||
+ | - corpus of manually anotated coreference | ||
+ | - POS tagger | ||
+ | - coreference resolution procedure | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ===== Input Texts ===== | ||
+ | |||
+ | === Text Selection === | ||
+ | * CS data | ||
+ | * Anja's data ## // PDT data that are currently being annotated for the extended coreference // | ||
+ | * **JM**: It would be nice if the players could choose a domain of the texts to play on (science-fiction, | ||
+ | ***JM (6/3/09)**: Predelal jsem data pro playcoref, ted obsahuji jenom koreference mezi uzly s tagy N nebo P. Data jsou v adresari: ''/ | ||
+ | ***BH (11/ | ||
+ | * **EN** | ||
+ | * search the data that are available; **BH (11/3/09)** Z dokumentace dat, ktera bychom meli mit, jsem nasla MUC6, ale nevidim tam data s koreferenci. Jirka zjisti, jestli jsou nekde jinde nebo jak jinak se k nim muzeme dostat. | ||
+ | === Coding === | ||
+ | * utf-8 | ||
+ | |||
+ | === Internal format === | ||
+ | * sgml ## //propose dtd file: include the element '' | ||
+ | |||
+ | === (Pre)processing === | ||
+ | * tagging ## //see Tools needed below// | ||
+ | * acr by Linh ## // dtto // | ||
+ | |||
+ | === Text handling === | ||
+ | * sentence by sentence | ||
+ | * supervised selection of documents for a session | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ===== Scoring ===== | ||
+ | * '' | ||
+ | |||
+ | **JM**: | ||
+ | Já myslím, že do shody je tlačit chceme. Je žádoucí, aby anotace byla co nejúplnější. Když druhý hráč uvidí, že první hráč spojil nějaké slovo, vyvíjí to na něj tlak, aby se podíval, jestli to | ||
+ | nepřehlédl a jestli by ho nemohl zapojit také. Neukazuje se mu kam, takže když nenajde žádný cíl, nezapojí ho a bude se radovat, že první hráč udělal nějakou chybu. Myslím, že ta funkce by měla brát **buď** automatickou anotaci **nebo** manuální, podle toho, co je k dispozici. Rovněž si teď myslím, že manuálně anotovaná data budeme používat minimálně - pouze pro změření úspěšnosti anotace pomocí hry - to ale nemusí být vůbec součástí skóre hry, to se udělá off-line. Manuálně anotovaných dat máme málo, jsou už anotovaná a nejsou zábavná. Z toho mi vyplývá, že bych manuální anotaci pro určování skóre nebral vůbec v úvahu a ze vzorečku nahoře bych první člen vyhodil. | ||
+ | |||
+ | **BH**: Jirka ma pravdu. Pocitani skore musi byt objektivni. Proto jsem vzorecek upravila tak, ze nebude pocitat shodu hrace vzhledem k rucni anotaci. | ||
+ | |||
+ | |||
+ | |||
+ | ===== Output Data Needed ===== | ||
+ | * score list ## // | ||
+ | * documents after the '' | ||
+ | (**JM**: Mluvil jsem kvůli měření mezianotátorské shody v anotování koreference se Zdeňkem a vyšlo z toho, že na měření shody na šipkách by použil prostě jen F-measure. Její smysl je jasný a je symetrická. Kappa je nevhodná kvůli tomu, že pravděpodobnost náhodné shody je poměrně nízká a těžko se určuje; kappa se hodí spíš pro klasifikační úlohy (proto ji použiju v Anjiině projektu na shodu v určování typu koreference, | ||
+ | - kappa measure | ||
+ | - G-theory - see [[http:// | ||
+ | Identifying Sources of Disagreement: | ||
+ | - the Pearson correlation - see (Snow et al., 2008) [[http:// | ||
+ | * session | ||
+ | * player_A_id, | ||
+ | * document(s) | ||
+ | * number of corrections by player_A and by player_B (**JM**: I do not see the point in this) | ||
+ | * corrections by player_A and by player_B (**JM**: and maybe nor in this) (**BH**: I am interested in the players' | ||
+ | |||
+ | |||
+ | ===== Design ===== | ||
+ | * What info to be displayed in the session? | ||
+ | * session time = elapsed time + remaining time | ||
+ | * how many sentences my partner has read so far | ||
+ | * running pts **???????** (**JM**: I would be very cautious with this; the user might be tempted to cancel an action if the score decreases; the user might also try to fit the automatic annotation (by trying various arrows and watching if the score goes up or down), which is not what we want) | ||
+ | * Format of the text | ||
+ | * **JM**: nouns and pronouns might be displayed slightly differently so that the user avoids other parts of speech easily; he should not be allowed to use other parts of speech at all | ||
+ | * Visualization of the coreference pairs | ||
+ | * colors | ||
+ | * arrows (**JM**: to avoid too many arrows on the screen, possibly only if the mouse pointer hovers over a word, arrows that start or end at the word would be displayed) | ||
+ | * ... | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ===== Tools needed ===== | ||
+ | * tagger ## tool_chain (CAC2.0) | ||
+ | * Linh's coreference resolution procedure - see TectoMT - **JM** | ||
+ | * vyzkouset - trenink a test - na datech Anji | ||
+ | * conversion: csts <-> pml m_coref scheme | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ====== ACL - IJCNLP2009 ====== | ||
+ | * [[http:// | ||
+ | * [[http:// | ||
+ | * pracovni adresar ''/ |