TREC 2023 Tip-of-the-Tongue (ToT) Track

Overview paper: trec.nist.gov/pubs/trec32/papers/Overview_tot.pdf
Overview presentation: docs.google.com/presentation/d/1gMToQywybkU3ajdF4…
Track proceedings: trec.nist.gov/pubs/trec32/xref.html#tot

Note: You are viewing the guidelines for the 2023 edition of the TREC ToT track. Please see the latest guidelines for the upcoming / latest edition of the track.

Guidelines

Important dates

Registration

To participate in TREC please pre-register at the following website: https://ir.nist.gov/trecsubmit.open/application.html.

Task definition

In terms of input and output, the movie identification task is relatively straightforward—given an input TOT request, output a ranked list of movies. Each movie must be identified by its Wikipedia page id and the correct movie should be ranked as high as possible. For each query, runs should return a ranked list of 1000 Wikipedia page ids. Runs will be evaluated using IR metrics that are appropriate for IR tasks with one relevant document, such as discounted cumulative gain, reciprocal rank, and success@k.

Datasets

Data can be downloaded using the links below. Participants can download the files in a single zip file (TREC-ToT.zip) or download each file (corpus.jsonl.zip, train.zip, dev.zip) separately. See Corpora and Queries for a description of the files (train/dev folders also include qrel files in the TREC format).

Description Link MD5 Hash # entries
corpus, train and dev set (single zip) TREC-ToT.zip f84fe82cb80e3ee1072576c8d6c4a417 -
corpus (JSONL) corpus.jsonl.zip 3bdf2941d256c88d1fbfd9dbc483f43d 231618
train queries (JSONL) & qrel.txt train.zip f5e7e07e409f214b5a67af7e23b231c7 150
dev queries (JSONL) & qrel.txt dev.zip 40ef6eace079a40f23237005db1b0f1a 150
test queries (JSONL) test.zip 756b77a2725f0960defb3a9b12d95334 150
Mapping: Wikipedia Page to IMDb ID wikipage_id_to_imdb.json.zip c837316e9374b577974dc14b12a15a2c 190370

Corpora

Wikipedia Corpus: Subset of Wikipedia pages directly or indirectly associated with (relevant sub-classes of) the “audiovisual works” category (231,852 pages). This corpus contains the relevant movie for all TOT queries in the training, development, and test sets described below. In addition to the fields described below, participants can utilize this file to obtain doc_id to IMDb identifier(s). Note that only some (190370 of 231618) pages have an associated ID.

Each document in the corpus will be described by the following fields:

An example document is described below.

Example Document

{
   "page_title":"Actrius",
   "page_source": "{{Use dmy dates|date=March 2021}}\n{{Infobox film\n| name           = Actresses\n| image          = Actrius film poster.jpg\n| alt            = \n| caption        = Catalan language film poster\n| native_name      = ([[Catalan language|Catalan]]: '''''Actrius''''')\n| director       = [[Ventura Pons]]\n| producer       = Ventura Pons\n| writer         = [[Josep Maria Benet i Jornet]]\n| screenplay     = Ventura Pons\n| story          = \n| based_on       = {{based on|(stage play) ''E.R.''|Josep Maria Benet i Jornet}}\n| starring       = {{ubl|[[N\u00faria Espert]]|[[Rosa Maria Sard\u00e0]]|[[Anna Lizaran]]|[[Merc\u00e8 Pons]]}}\n| narrator       = <!-- or: |narrators = -->\n| music          = Carles Cases\n| cinematography = Tom\u00e0s Pladevall\n| editing        = Pere Abadal\n| production_companies = {{ubl|[[Canal+|Canal+ Espa\u00f1a]]|Els Films de la Rambla S.A.|[[Generalitat de Catalunya|Generalitat de Catalunya - Departament de Cultura]]|[[Televisi\u00f3n Espa\u00f1ola]]}}\n| distributor    = [[Buena Vista International]]\n| released       = {{film date|df=yes|1997|1|17|[[Spain]]}}\n| runtime        = 100 minutes\n| country        = Spain\n| language       = Catalan\n| budget         = \n| gross          = <!--(please use condensed and rounded values, e.g. \"\u00a311.6 million\" not \"\u00a311,586,221\")-->\n}}\n\n'''''Actresses''''' ([[Catalan language|Catalan]]: '''''Actrius''''') is a 1997 [[Catalan language]] Spanish drama film produced and directed by [[Ventura Pons]] and based on the award-winning stage play ''E.R.'' by [[Josep Maria Benet i Jornet]]. The film has no male actors, with all roles played by females.<ref name=\"El Pais\">{{cite news|last1=Torres|first1=Rosanna|title='E. R', de Benet i Jornet, es llevada al cine y al teatro|url=http://elpais.com/diario/1996/10/15/cultura/845330405_850215.html|access-date=2015-12-21|lang=es|work=[[El Pa\u00eds]]|date=1996-10-15|archive-url=https://web.archive.org/web/20121014165502/http://elpais.com/diario/1996/10/15/cultura/845330405_850215.html|archive-date=2012-10-14}}</ref> The film was produced in 1996. <ref>{{cite web |title=Actrius |url=https://www.bcncatfilmcommission.com/en/films/actrius|lang=en|website=Barcelona Film Commission |access-date=2021-05-13|archive-url=https://web.archive.org/web/20210513150105/https://www.bcncatfilmcommission.com/en/films/actrius|archive-date=2021-05-13|url-status=live}}</ref>\n\n== Synopsis ==\nIn order to prepare herself to play a role commemorating the life of legendary actress Empar Ribera, young actress ([[Merc\u00e8 Pons]]) interviews three established actresses who had been the Ribera's pupils: the international diva Gl\u00f2ria Marc ([[N\u00faria Espert]]), the television star Assumpta Roca ([[Rosa Maria Sard\u00e0]]), and dubbing director Maria Caminal ([[Anna Lizaran]]).<ref name=SFF>{{cite web | url=http://www.stockholmfilmfestival.se/en/festival/1997/film/actrius | title=Actrius |lang=en|publisher=[[Stockholm International Film Festival]] |date=1997 | access-date=2015-12-21|archive-url=https://web.archive.org/web/20151222175333/http://www.stockholmfilmfestival.se/en/festival/1997/film/actrius|archive-date=2015-12-22|url-status=dead }}</ref>\n\n== Cast ==\n* [[N\u00faria Espert]] as Gl\u00f2ria Marc\n* [[Rosa Maria Sard\u00e0]] as Assumpta Roca\n* [[Anna Lizaran]] as Maria Caminal\n* [[Merc\u00e8 Pons]] as Estudiant\n\n== Recognition ==\n\n=== Screenings ===\n''Actrius'' screened in 2001 at the [[Grauman's Egyptian Theatre]] in an [[American Cinematheque]] retrospective of the works of its director. The film had first screened at the same location in 1998.<ref name=\"LA Times\">{{cite news|last1=THomas|first1=Kevin|title=Sometimes, the World Gets in the Way|url=http://articles.latimes.com/2001/mar/01/entertainment/ca-31570|lang=en|access-date=2015-12-21|newspaper=[[Los Angeles Times]]|date=2001-03-01|archive-url=https://web.archive.org/web/20121221084433/http://articles.latimes.com/2001/mar/01/entertainment/ca-31570|archive-date=2012-12-21|url-status=live}}</ref> It was also shown at the 1997 [[Stockholm International Film Festival]].<ref name=SFF />\n\n=== Reception ===\nIn ''Movie - Film - Review'', Christopher Tookey wrote that though the actresses were \"competent in roles that may have some reference to their own careers\", the film \"is visually unimaginative, never escapes its stage origins, and is almost totally lacking in revelation or surprising incident\".<ref name=Tookey/> Noting that there were \"occasional, refreshing moments of intergenerational bitchiness\", they did not \"justify comparisons to ''[[All About Eve]]''\", and were \"insufficiently different to deserve critical parallels with ''[[Rashomon]]''\".<ref name=Tookey/> He also wrote that ''[[The Guardian]]'' called the film a \"slow, stuffy chamber-piece\", and that ''[[The Evening Standard]]'' stated the film's \"best moments exhibit the bitchy tantrums seething beneath the threesome's composed veneers\".<ref name=Tookey>{{cite web|last1=Tookey|first1=Chris|title=review: Actresses / Actrius / Actrices|url=http://www.movie-film-review.com/devFilm.asp?ID=12423|lang=en|publisher=Movie - Film - Review|access-date=2015-12-21|archive-url=https://web.archive.org/web/20120501211137/http://www.movie-film-review.com/devFilm.asp?ID=12423|archive-date=2012-05-01|url-status=dead}}</ref> [[MRQE]] wrote \"This cinematic adaptation of a theatrical work is true to the original, but does not stray far from a theatrical rendering of the story.\"<ref name=MRQE>{{cite web|last1=staff|title=Actrius (1997)|url=http://www.mrqe.com/movie_reviews/actrius-m100030469|lang=en|publisher=[[MRQE]]|access-date=2015-12-21|archive-url=https://web.archive.org/web/20151222150358/http://www.mrqe.com/movie_reviews/actrius-m100030469|archive-date=2015-12-22}}</ref>\n\n=== Awards and nominations ===\n* 1997, won 'Best Catalan Film' at [[Butaca Awards]] for [[Ventura Pons]]\n* 1997, won 'Best Catalan Film Actress' at Butaca Awards, shared by [[N\u00faria Espert]], [[Rosa Maria Sard\u00e0]], [[Anna Lizaran]], and [[Merc\u00e8 Pons]]\n* 1998, nominated for 'Best Screenplay' at [[Goya Awards]], shared by [[Josep Maria Benet i Jornet]] and Ventura Pons\n\n== References ==\n{{reflist}}\n\n== External links ==\n* {{IMDb title|0115462|Actresses}}\n* {{official website|https://web.archive.org/web/20090217140746/http://venturapons.com/filmografia/actrices.html}} [[Wayback Machine|as archived 17 February 2009]] (Spanish)\n\n[[Category:1997 films]]\n[[Category:1997 drama films]]\n[[Category:Catalan-language films]]\n[[Category:Films set in Barcelona]]\n[[Category:Films directed by Ventura Pons]]\n[[Category:Spanish drama films]]\n[[Category:1990s Spanish films]]" , 
   "wikidata_id":"Q2823770",
   "wikidata_classes":[
      [
         "Q11424",
         "film"
      ]
   ],
   "text":"Actresses (Catalan: Actrius) is a 1997 Catalan language Spanish drama film produced and directed by Ventura Pons and based on the award-winning stage play \"E.R.\" by Josep Maria Benet i Jornet. The film has no male actors, with all roles played by females. The film was produced in 1996. \nSynopsis.\nIn order to prepare herself to play a role commemorating the life of legendary actress Empar Ribera, young actress (Merc\u00e8 Pons) interviews three established actresses who had been the Ribera's pupils: the international diva Gl\u00f2ria Marc (N\u00faria Espert), the television star Assumpta Roca (Rosa Maria Sard\u00e0), and dubbing director Maria Caminal (Anna Lizaran).\nRecognition.\nScreenings.\n\"Actrius\" screened in 2001 at the Grauman's Egyptian Theatre in an American Cinematheque retrospective of the works of its director. The film had first screened at the same location in 1998. It was also shown at the 1997 Stockholm International Film Festival.\nReception.\nIn \"Movie - Film - Review\", Christopher Tookey wrote that though the actresses were \"competent in roles that may have some reference to their own careers\", the film \"is visually unimaginative, never escapes its stage origins, and is almost totally lacking in revelation or surprising incident\". Noting that there were \"occasional, refreshing moments of intergenerational bitchiness\", they did not \"justify comparisons to \"All About Eve\"\", and were \"insufficiently different to deserve critical parallels with \"Rashomon\"\". He also wrote that \"The Guardian\" called the film a \"slow, stuffy chamber-piece\", and that \"The Evening Standard\" stated the film's \"best moments exhibit the bitchy tantrums seething beneath the threesome's composed veneers\". MRQE wrote \"This cinematic adaptation of a theatrical work is true to the original, but does not stray far from a theatrical rendering of the story.\"",
   "sections":{
      "abstract":"Actresses (Catalan: Actrius) is a 1997 Catalan language Spanish drama film produced and directed by Ventura Pons and based on the award-winning stage play \"E.R.\" by Josep Maria Benet i Jornet. The film has no male actors, with all roles played by females. The film was produced in 1996. ",
      "synopsis":"Synopsis.\nIn order to prepare herself to play a role commemorating the life of legendary actress Empar Ribera, young actress (Merc\u00e8 Pons) interviews three established actresses who had been the Ribera's pupils: the international diva Gl\u00f2ria Marc (N\u00faria Espert), the television star Assumpta Roca (Rosa Maria Sard\u00e0), and dubbing director Maria Caminal (Anna Lizaran).",
      "recognition":"Recognition.\nScreenings.\n\"Actrius\" screened in 2001 at the Grauman's Egyptian Theatre in an American Cinematheque retrospective of the works of its director. The film had first screened at the same location in 1998. It was also shown at the 1997 Stockholm International Film Festival.\nReception.\nIn \"Movie - Film - Review\", Christopher Tookey wrote that though the actresses were \"competent in roles that may have some reference to their own careers\", the film \"is visually unimaginative, never escapes its stage origins, and is almost totally lacking in revelation or surprising incident\". Noting that there were \"occasional, refreshing moments of intergenerational bitchiness\", they did not \"justify comparisons to \"All About Eve\"\", and were \"insufficiently different to deserve critical parallels with \"Rashomon\"\". He also wrote that \"The Guardian\" called the film a \"slow, stuffy chamber-piece\", and that \"The Evening Standard\" stated the film's \"best moments exhibit the bitchy tantrums seething beneath the threesome's composed veneers\". MRQE wrote \"This cinematic adaptation of a theatrical work is true to the original, but does not stray far from a theatrical rendering of the story.\"",
      "screenings":"Screenings.\n\"Actrius\" screened in 2001 at the Grauman's Egyptian Theatre in an American Cinematheque retrospective of the works of its director. The film had first screened at the same location in 1998. It was also shown at the 1997 Stockholm International Film Festival.",
      "reception":"Reception.\nIn \"Movie - Film - Review\", Christopher Tookey wrote that though the actresses were \"competent in roles that may have some reference to their own careers\", the film \"is visually unimaginative, never escapes its stage origins, and is almost totally lacking in revelation or surprising incident\". Noting that there were \"occasional, refreshing moments of intergenerational bitchiness\", they did not \"justify comparisons to \"All About Eve\"\", and were \"insufficiently different to deserve critical parallels with \"Rashomon\"\". He also wrote that \"The Guardian\" called the film a \"slow, stuffy chamber-piece\", and that \"The Evening Standard\" stated the film's \"best moments exhibit the bitchy tantrums seething beneath the threesome's composed veneers\". MRQE wrote \"This cinematic adaptation of a theatrical work is true to the original, but does not stray far from a theatrical rendering of the story.\""
   },
   "infoboxes":[
      {
         "name":"film",
         "params":{
            "name":"Actresses",
            "image":"Actrius film poster.jpg",
            "alt":"",
            "caption":"Catalan language film poster",
            "native_name":"([[Catalan language|Catalan]]: '''''Actrius''''')",
            "director":"[[Ventura Pons]]",
            "producer":"Ventura Pons",
            "writer":"[[Josep Maria Benet i Jornet]]",
            "screenplay":"Ventura Pons",
            "story":"",
            "based_on": "{{based on|(stage play) ''E.R.''|Josep Maria Benet i Jornet}}",
            "starring":"{{ubl|[[N\u00faria Espert]]|[[Rosa Maria Sard\u00e0]]|[[Anna Lizaran]]|[[Merc\u00e8 Pons]]}}",
            "narrator":"<!-- or: |narrators = -->",
            "music":"Carles Cases",
            "cinematography":"Tom\u00e0s Pladevall",
            "editing":"Pere Abadal",
             "production_companies":"{{ubl|[[Canal+|Canal+ Espa\u00f1a]]|Els Films de la Rambla S.A.|[[Generalitat de Catalunya|Generalitat de Catalunya - Departament de Cultura]]|[[Televisi\u00f3n Espa\u00f1ola]]}}", 
            "distributor":"[[Buena Vista International]]",
             "released":"{{film date|df=yes|1997|1|17|[[Spain]]}}", 
            "runtime":"100 minutes",
            "country":"Spain",
            "language":"Catalan",
            "budget":"",
            "gross":"<!--(please use condensed and rounded values, e.g. \"\u00a311.6 million\" not \"\u00a311,586,221\")-->"
         }
      }
   ],
   "doc_id":"330"
}

Queries

Queries (or topics in TREC lingo) are sourced from the MS-TOT dataset. Participating groups will be given a JSONL file consisting of a random sample of 150 MS-TOT queries each for training, development, and test. An example query is described below. Corresponding to these queries, sentence-level annotations will also be distributed. Participating groups are encouraged to leverage these codes however they want. This might include treating sentences associated with specific codes differently (e.g., ignoring or down-weighing sentences that convey uncertainty).

Example query

Note: Only a single sentence annotation is shown in the example below


{
   "id":"763",
   "url":"https://irememberthismovie.com/super-rare-surreal-dystopian-masterpiece/",
   "domain":"movie",
   "title":"Super Rare Surreal Dystopian Masterpiece",
   "text":"Very rare movie that is scifi/dystopian/experimental/surreal. It\u2019s like Stalker meets el Topo meets Holy Mountain meets Alphaville meets Delicatessen meets Hard to be a God, like Kurosawa, Tarkovsky, and Lynch had a kid together. It was color, possibly Russian, and I don\u2019t really remember the decade but want to say 60s or 70s, though could easily be more recent. It is VERY rare, there is only one crappy partial print of it, and that is what the youtube version is from. Lot of wide shots in a surreal wilderness, winter settings, strange bleeding saturation in some shots. Crazy costumes. Seriously one of the strangest films I\u2019ve ever seen and my favorite films are strange/weird ones. If you\u2019ve ever seen what you\u2019re thinking of on a \u201cbest weird movies\u201d or \u201cyou\u2019ve never seen this!\u201d list, that\u2019s NOT it. I don\u2019t think this film even has a cult following of ten people. It\u2019s an actual rare gem. Have been looking through selections at 366 Weird Movies and not found it yet (btw the way most of those titles are exactly the kind of not-actually-rare movies this film is definitely not).",
   "wikipedia_id":"16742289",
   "wikipedia_url":"https://en.wikipedia.org/wiki/On_the_Silver_Globe_(film)",
   "wikidata_id":"Q1988165",
   "imdb_url":"https://www.imdb.com/title/tt0093593",
   "sentence_annotations":[
      {
         "id":1,
         "text":"Very rare movie that is scifi/dystopian/experimental/surreal.",
         "labels":{
            "opinion":false,
            "emotion":false,
            "hedging":true,
            "social":false,
            "comparison_relative":false,
            "search":false,
            "movie":{
               "music_compare":false,
               "genre_audience":false,
               "production_camera_angle":false,
               "timeframe_singular":false,
               "origin_actor":false,
               "object":false,
               "location_specific":false,
               "person_fictional":false,
               "production_visual":false,
               "scene":false,
               "negation":false,
               "production_audio":false,
               "character":false,
               "genre_traditional_tone":true,
               "release_date":false,
               "location_type":false,
               "origin_language":false,
               "origin_movie":false,
               "person_real":false,
               "plot":false,
               "quote":false,
               "category":true,
               "music_specific":false,
               "timeframe_plural":false
            },
            "context":{
               "cross_media":false,
               "physical_medium":false,
               "situational_count":false,
               "physical_user_location":false,
               "situational_evidence":false,
               "situational_witness":false,
               "temporal":false
            }
         }
      }
   ]
}

Sentence annotations codes

Use of external information

You are generally allowed to use external information while developing your runs. When you submit your runs, please fill in a form listing what resources you used. This could include an external corpus such as TMDB or a pretrained model (e.g. word embeddings, BERT). Additionally,

Training data documentation and declaration

The emerging practices of pretraining large language models on web-scale datasets and community sharing of these pretrained models for downstream training / model development creates challenges in tracking what training data have been used for particular run submissions. Given that our test queries are sampled from the MS-TOT dataset that has been publicly available online since 2021, there is a possibility that this data may have been used for training models that are employed in submitted runs. To ensure that we get robust scientific conclusions and insights from this year’s track, we are requesting all participants to document all training data that were employed in the preparation of a given run to the best of your knowledge (including any data used for pretraining models that you are building on top of). In addition, we are also requesting participants to declare if they are 100% confident that no data from https://github.com/microsoft/Tip-of-the-Tongue-Known-Item-Retrieval-Dataset-for-Movie-Identification or iRememberThisMovie.com was used for training. We recognize that not all pretrained models publicly document their training data in details. So, please mark that declaration to be true only if you are 100% confident about all training data used in preparation of your run and can guarantee that it does not include any data from the sources mentioned. We do not discourage submissions where there may be some uncertainty about the training data but we want to be aware of this at the time of analyzing the evaluated results.

Baselines

We’ve released three baselines, BM25, a dense retriever and GPT-4, in this repository. A summary of baseline results, along with corresonding run files are provided below.:

Benchmark Runfiles Dev-DCG Dev-Success@1000 Dev-MRR
BM25 (k1=0.8, b=1.0) train, dev, test 0.1314 0.4067 0.0881
Dense Retrieval (SBERT) (Distilbert) train, dev, test 0.1627 0.6600 0.0743
GPT-4 train, dev, test 0.2407 0.3200 0.2180

Submission and evaluation

Submission form: https://ir.nist.gov/trecsubmit/tot.html.

We will be following a similar format as the ones used by most TREC submissions, which is repeated below. White space is used to separate columns. The width of the columns in the format is not important, but it is important to have exactly six columns per line with at least one space between the columns.

1 Q0 pid1    1 2.73 runid1
1 Q0 pid2    2 2.71 runid1
1 Q0 pid3    3 2.61 runid1
1 Q0 pid4    4 2.05 runid1
1 Q0 pid5    5 1.89 runid1

, where:

The main type of TREC submission is automatic, which means there was not manual intervention in running the test queries. This means you should not adjust your runs, rewrite the query, retrain your model, or make any other sorts of manual adjustments after you see the test queries. The ideal case is that you only look at the test queries to check that they ran properly (i.e. no bugs) then you submit your automatic runs. However, if you want to have a human in the loop for your run, or do anything else that uses the test queries to adjust your model or ranking, you can mark your run as manual. Manual runs are interesting, and we may learn a lot, but these are distinct from our main scenario which is a system that responds to unseen queries automatically.

Runs will be evaluated using metrics appropriate for retrieval scenarios with one relevant document. In particular, our primary evaluation metric for this year’s track will be discounted cumulative gain (DCG) but we may also compute other metrics such as reciprocal rank (RR) and success@k.