Bart Large Cnn Samsum

🚀 英語会話要約モデル

このモデルは英語の会話を要約するためのものです。事前学習はBART-large-cnnをベースに行われています。実装と微調整のスクリプトはこちらで確認できます: https://github.com/joceyngan/conversation_summarization

📄 ライセンス

このプロジェクトはMITライセンスの下で公開されています。

📦 使用データセット

• Samsung/samsum

🚀 クイックスタート

評価結果

参照要約はデータセットの正解データで、生成要約はこのモデルによって生成されたものです。

サンプル1

原文

A: Hi Tom, are you busy tomorrow’s afternoon? B: I’m pretty sure I am. What’s up? A: Can you go with me to the animal shelter?. B: What do you want to do? A: I want to get a puppy for my son. B: That will make him so happy. A: Yeah, we’ve discussed it many times. I think he’s ready now. B: That’s good. Raising a dog is a tough issue. Like having a baby ;-) A: I'll get him one of those little dogs. B: One that won't grow up too big;-) A: And eat too much;-)) B: Do you know which one he would like? A: Oh, yes, I took him there last Monday. He showed me one that he really liked. B: I bet you had to drag him away. A: He wanted to take it home right away ;-). B: I wonder what he'll name it. A: He said he’d name it after his dead hamster – Lemmy - he's a great Motorhead fan :-)))

参照要約

A will go to the animal shelter tomorrow to get a puppy for her son. They already visited the shelter last Monday and the son chose the puppy.

生成要約

A wants to get a puppy for his son. B will go with him to the animal shelter tomorrow. A will get him one of those little dogs. A took him there last Monday and he liked it a lot. A wants to name it Lemmy.

サンプル2

原文

Emma: I’ve just fallen in love with this advent calendar! Awesome! I wanna one for my kids! Rob: I used to get one every year as a child! Loved them! Emma: Yeah, i remember! they were filled with chocolates! Lauren: they are different these days! much more sophisticated! Haha! Rob: yeah, they can be fabric/ wooden, shop bought/ homemade, filled with various stuff Emma: what do you fit inside? Lauren: small toys, Christmas decorations, creative stuff, hair bands & clips, stickers, pencils & rubbers, small puzzles, sweets Emma: WOW! That’s brill! X Lauren: i add one more very special thing as well- little notes asking my children to do something nice for someone else Rob: i like that! My sister adds notes asking her kids questions about christmas such as What did the 3 wise men bring? etc Lauren: i reckon it prepares them for Christmas Emma: and makes it more about traditions and being kind to other people Lauren: my children get very excited every time they get one! Emma: i can see why! :)

参照要約

Emma and Rob love the advent calendar. Lauren fits inside calendar various items, for instance, small toys and Christmas decorations. Her children are excited whenever they get the calendar.

生成要約

Emma loves the advent calendar. Rob used to get one every year as a child. Emma would like to buy one for her kids. Rob's sister asks her kids questions about Christmas. Lauren's children get very excited every time they get one.

サンプル3

原文

Jackie: Madison is pregnant Jackie: but she doesn't wanna talk about it Iggy: why Jackie: I don't know why because she doesn't wanna talk about it Iggy: ok Jackie: I wanted to prepare you for it because people get super excited and ask lots of questions Jackie: and she looked way more anxious than excited Iggy: she's probably worrying about it Iggy: she's taking every commitment really seriously Jackie: it could be money problems or relationship problems Iggy: or maybe she wants an abortion Jackie: it could be all of the above Iggy: but you know what? Iggy: once my friend was pregnant and I couldn't bring myself to be happy about it Jackie: why? Iggy: I felt they were immature and I couldn't picture this couple as parents Jackie: I felt similar way on Patricia's wedding Iggy: Patricia Stevens? Jackie: yes Iggy: so we're talking about the same person Jackie: what a coincidence Jackie: so she's pregnant? Iggy: she thought she was Jackie: damn...

参照要約

Madison is pregnant but she doesn't want to talk about it. Patricia Stevens got married and she thought she was pregnant.

生成要約

Madison is pregnant. She doesn't want to talk about it. Iggy's friend Patricia Stevens was pregnant with Patricia Stevens and she didn't like it much. She felt they were immature and she couldn't picture them as parents.

サンプル4

原文

Marla: <file_photo> Marla: look what I found under my bed Kiki: lol Tamara: is that someone's underwear? Marla: it certainly isn't mine, my ass is big but it isn't huge Kiki: it looks like male underwear Tamara: not necessarily, maybe some butch had fun in your room while you were gone Marla: ok but how can you leave your underwear after hooking up? wtf is wrong with people Kiki: she or he could be too wasted to notice Tamara: or maybe someone put their pants there to piss you off Marla: that makes no sense Marla: it's so fucking childish Kiki: if it's childish then it must have been your sister's idea Marla: she's 13, she doesn't have underwear that isn't pink Tamara: maybe it belonged to one of your exes? Kiki: she would have recognized it Marla: lol we're doing total CSI investigation on one pair of boxers :D Kiki: <file_gif> Tamara: lol Tamara: I think your sister convinced someone to put their underwear in your room as a dare Marla: sounds legit Kiki: Tamara, you just cracked the case! Tamara: <file_gif> Tamara: always happy to help

参照要約

Marla found a pair of boxers under her bed.

生成要約

Marla found a pair of boxers under her bed. Kiki, Tamara, Marla and Tamara are laughing at the fact that someone left their underwear under Marla's bed after hooking up. Marla is convinced that her sister convinced someone to put their underwear in her room.

サンプル5

原文

Robert: Hey give me the address of this music shop you mentioned before Robert: I have to buy guitar cable Fred: <file_other> Fred: Catch it on google maps Robert: thx m8 Fred: ur welcome

参照要約

Robert wants Fred to send him the address of the music shop as he needs to buy guitar cable.

生成要約

Fred gives Robert the address of the music shop where he needs to buy guitar cable. Robert can find it on google maps. Robert has to buy a guitar cable at the shop. Fred sends him a link to the address.

微調整引数

{
    output_dir=str(results_path),
    evaluation_strategy="epoch",
    save_strategy="epoch",
    logging_steps=10,
    learning_rate=2e-5,
    per_device_train_batch_size=2,
    per_device_eval_batch_size=2,
    num_train_epochs=3,
    weight_decay=0.01,
    report_to="tensorboard",
    save_total_limit=3,
    load_best_model_at_end=True,
    metric_for_best_model="eval_loss",
    greater_is_better=False,
    logging_dir=str(results_path),
}

使用方法

tokenizer = BartTokenizer.from_pretrained('facebook/bart-large-cnn')
model = BartForConditionalGeneration.from_pretrained('jngan/bart-large-cnn-samsum')