import gradio as gr import os from transformers import AutoTokenizer, AutoModel from sentence_transformers import SentenceTransformer import pickle import nltk nltk.download('punkt') # tokenizer nltk.download('averaged_perceptron_tagger') # postagger import time from input_format import * from score import * # load document scoring model torch.cuda.is_available = lambda : False device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') pretrained_model = 'allenai/specter' tokenizer = AutoTokenizer.from_pretrained(pretrained_model) doc_model = AutoModel.from_pretrained(pretrained_model) doc_model.to(device) # load sentence model sent_model = SentenceTransformer('sentence-transformers/gtr-t5-base') sent_model.to(device) def get_similar_paper( abstract_text_input, pdf_file_input, author_id_input, num_papers_show=10 ): print('retrieving similar papers...') start = time.time() input_sentences = sent_tokenize(abstract_text_input) # TODO handle pdf file input if pdf_file_input is not None: name = None papers = [] raise ValueError('Use submission abstract instead.') else: # Get author papers from id name, papers = get_text_from_author_id(author_id_input) # Compute Doc-level affinity scores for the Papers print('computing document scores...') # TODO detect duplicate papers? titles, abstracts, doc_scores = compute_document_score( doc_model, tokenizer, abstract_text_input, papers, batch=50 ) tmp = { 'titles': titles, 'abstracts': abstracts, 'doc_scores': doc_scores } # Select top K choices of papers to show titles = titles[:num_papers_show] abstracts = abstracts[:num_papers_show] doc_scores = doc_scores[:num_papers_show] display_title = ['[ %0.3f ] %s'%(s, t) for t, s in zip(titles, doc_scores)] end = time.time() print('paper retrieval complete in [%0.2f] seconds'%(end - start)) print('obtaining highlights..') start = time.time() input_sentences = sent_tokenize(abstract_text_input) num_sents = len(input_sentences) summary_info = dict() # elements to visualize upfront for aa, (tt, ab, ds) in enumerate(zip(titles, abstracts, doc_scores)): # Compute sent-level and phrase-level affinity scores for each papers sent_ids, sent_scores, info, top_pairs_info = get_highlight_info( sent_model, abstract_text_input, ab, K=2 ) # get scores for each word in the format for Gradio Interpretation component word_scores = dict() for i in range(num_sents): word_scores[str(i)] = { "original": ab, "interpretation": list(zip(info['all_words'], info[i]['scores'])) } tmp[display_title[aa]] = { 'title': tt, 'abstract': ab, 'doc_score': ds, 'source_sentences': input_sentences, 'highlight': word_scores, 'top_pairs': top_pairs_info } # TODO better ways of saving intermediate results? user identifiers per session? pickle.dump(tmp, open('info.pkl', 'wb')) end = time.time() print('done in [%0.2f] seconds'%(end - start)) # set up elements to show out = [ gr.update(choices=display_title, interactive=True, visible=False), # set of papers (radio) gr.update(choices=input_sentences, interactive=True, visible=False) # submission sentences ] # set up elements to visualize upfront top_papers_show = 3 # number of top papers to show upfront top_num_info_show = 2 # number of sentence pairs from each paper to show upfront summary_out = [] for i in range(top_papers_show): out_tmp = [ gr.update(value=titles[i], visible=True), gr.update(value=doc_scores[i], visible=True) ] tp = tmp[display_title[i]]['top_pairs'] for j in range(top_num_info_show): out_tmp += [ gr.update(value=tp[j]['score'], visible=True), tp[j]['query']['original'], tp[j]['query'], tp[j]['candidate']['original'], tp[j]['candidate'] ] summary_out += out_tmp # add updates to the show more button out = out + summary_out + [gr.update(visible=True)] # show more button assert(len(out) == (top_num_info_show * 5 + 2) * top_papers_show + 3) return tuple(out) def show_more(): return ( gr.update(visible=True), # set of papers gr.update(visible=True), # submission sentences gr.update(visible=True), # title row gr.update(visible=True), # abstract row ) def update_name(author_id_input): # update the name of the author based on the id input name, _ = get_text_from_author_id(author_id_input) return gr.update(value=name) def change_output_highlight(selected_papers_radio, source_sent_choice): # change the output highlight based on the sentence selected from the submission fname = 'info.pkl' if os.path.exists(fname): tmp = pickle.load(open(fname, 'rb')) source_sents = tmp[selected_papers_radio]['source_sentences'] highlights = tmp[selected_papers_radio]['highlight'] for i, s in enumerate(source_sents): #print('changing highlight') if source_sent_choice == s: return highlights[str(i)] else: return def change_paper(selected_papers_radio): # change the paper to show based on the paper selected fname = 'info.pkl' if os.path.exists(fname): tmp = pickle.load(open(fname, 'rb')) title = tmp[selected_papers_radio]['title'] abstract = tmp[selected_papers_radio]['abstract'] aff_score = tmp[selected_papers_radio]['doc_score'] highlights = tmp[selected_papers_radio]['highlight'] return title, abstract, aff_score, highlights['0'] else: return with gr.Blocks() as demo: # Text description about the app and disclaimer ### TEXT Description # TODO add instruction video link # TODO udpate instruction based on new changes gr.Markdown( """ # Paper Matching Helper This is a tool designed to help match an academic paper (submission) to a potential peer reviewer, by presenting information that may be relevant to the users. Below we describe how to use the tool. Also feel free to check out the [video]() for a more detailed rundown. ##### Input - The tool requires two inputs: (1) an academic paper's abstract in a text format, (2) and a potential reviewer's [Semantic Scholar](https://www.semanticscholar.org/) profile link. Once you put in a valid profile link, the reviewer's name will be displayed. - Once the name is confirmed, press the `What Makes this a Good Match?` button. ##### Similar Papers From the Reviewer - Based on the input information above, the tool will first search for similar papers from the reviewer's previous publications using [Semantic Scholar API](https://www.semanticscholar.org/product/api). - It will list top 10 similar papers along with the **affinity scores** (ranging from 0 -1) for each, computed using text representations from a [language model](https://github.com/allenai/specter/tree/master/specter). - You can click on different papers to see title, abstract, and affinity scores in detail. ##### Relevant Parts - Below the list of papers, we highlight relevant parts in the selected paper compared to the submission abstract. - On the left, you will see individual sentences from the submission abstract you can select from. - On the right, you will see the abstract of the selected paper, with **highlights**. - **Red highlights**: sentences from the reviewer's paper abstract with high semantic similarity to the selected sentence. - **Blue highlights**: phrases from the reviewer's paper abstract that is included in the selected sentence. - To see relevant parts in a different paper from the reviewer, select the new paper. ------- """ ) ### INPUT with gr.Row() as input_row: with gr.Column(): abstract_text_input = gr.Textbox(label='Submission Abstract') with gr.Column(): pdf_file_input = gr.File(label='OR upload a submission PDF File') with gr.Column(): with gr.Row(): author_id_input = gr.Textbox(label='Reviewer Link or ID (Semantic Scholar)') with gr.Row(): name = gr.Textbox(label='Confirm Reviewer Name', interactive=False) author_id_input.change(fn=update_name, inputs=author_id_input, outputs=name) with gr.Row(): compute_btn = gr.Button('What Makes This a Good Match?') ### OVERVIEW # Paper title, score, and top-ranking sentence pairs -- two sentence pairs per paper, three papers # TODO blockfy similar components together and simplify ## ONE BLOCK OF INFO FOR A SINGLE PAPER ## PAPER1 with gr.Row(): with gr.Column(scale=3): paper_title1 = gr.Textbox(label="From the reviewer's paper:", interactive=False, visible=False) with gr.Column(scale=1): affinity1 = gr.Number(label='Affinity', interactive=False, value=0, visible=False) with gr.Row() as rel1_1: with gr.Column(scale=1): sent_pair_score1_1 = gr.Number(label='Sentence Relevance', interactive=False, value=0, visible=False) with gr.Column(scale=4): sent_pair_source1_1 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_source1_1_hl = gr.components.Interpretation(sent_pair_source1_1) with gr.Column(scale=4): sent_pair_candidate1_1 = gr.Textbox(label='Sentence from Paper', visible=False) sent_pair_candidate1_1_hl = gr.components.Interpretation(sent_pair_candidate1_1) with gr.Row() as rel1_2: with gr.Column(scale=1): sent_pair_score1_2 = gr.Number(label='Sentence Relevance', interactive=False, value=0, visible=False) with gr.Column(scale=4): sent_pair_source1_2 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_source1_2_hl = gr.components.Interpretation(sent_pair_source1_2) with gr.Column(scale=4): sent_pair_candidate1_2 = gr.Textbox(label='Sentence from Paper', visible=False) sent_pair_candidate1_2_hl = gr.components.Interpretation(sent_pair_candidate1_2) ## PAPER 2 with gr.Row(): with gr.Column(scale=3): paper_title2 = gr.Textbox(label="From the reviewer's paper:", interactive=False, visible=False) with gr.Column(scale=1): affinity2 = gr.Number(label='Affinity', interactive=False, value=0, visible=False) with gr.Row() as rel2_1: with gr.Column(scale=1): sent_pair_score2_1 = gr.Number(label='Sentence Relevance', interactive=False, value=0, visible=False) with gr.Column(scale=4): sent_pair_source2_1 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_source2_1_hl = gr.components.Interpretation(sent_pair_source2_1) with gr.Column(scale=4): sent_pair_candidate2_1 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_candidate2_1_hl = gr.components.Interpretation(sent_pair_candidate2_1) with gr.Row() as rel2_2: with gr.Column(scale=1): sent_pair_score2_2 = gr.Number(label='Sentence Relevance', interactive=False, value=0, visible=False) with gr.Column(scale=4): sent_pair_source2_2 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_source2_2_hl = gr.components.Interpretation(sent_pair_source2_2) with gr.Column(scale=4): sent_pair_candidate2_2 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_candidate2_2_hl = gr.components.Interpretation(sent_pair_candidate2_2) ## PAPER 3 with gr.Row(): with gr.Column(scale=3): paper_title3 = gr.Textbox(label="From the reviewer's paper:", interactive=False, visible=False) with gr.Column(scale=1): affinity3 = gr.Number(label='Affinity', interactive=False, value=0, visible=False) with gr.Row() as rel3_1: with gr.Column(scale=1): sent_pair_score3_1 = gr.Number(label='Sentence Relevance', interactive=False, value=0, visible=False) with gr.Column(scale=4): sent_pair_source3_1 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_source3_1_hl = gr.components.Interpretation(sent_pair_source3_1) with gr.Column(scale=4): sent_pair_candidate3_1 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_candidate3_1_hl = gr.components.Interpretation(sent_pair_candidate3_1) with gr.Row() as rel3_2: with gr.Column(scale=1): sent_pair_score3_2 = gr.Number(label='Sentence Relevance', interactive=False, value=0, visible=False) with gr.Column(scale=4): sent_pair_source3_2 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_source3_2_hl = gr.components.Interpretation(sent_pair_source3_2) with gr.Column(scale=4): sent_pair_candidate3_2 = gr.Textbox(label='Sentence from Submission', visible=False) sent_pair_candidate3_2_hl = gr.components.Interpretation(sent_pair_candidate3_2) ## Show more button with gr.Row(): see_more_rel_btn = gr.Button('See more relevant parts from papers', visible=False) ### PAPER INFORMATION # show multiple papers in radio check box to select from with gr.Row(): selected_papers_radio = gr.Radio( choices=[], # will be udpated with the button click visible=False, # also will be updated with the button click label='Top Relevant Papers from the Reviewer' ) # selected paper information with gr.Row(visible=False) as title_row: with gr.Column(scale=3): paper_title = gr.Textbox(label='Title', interactive=False) with gr.Column(scale=1): affinity= gr.Number(label='Affinity', interactive=False, value=0) with gr.Row(): paper_abstract = gr.Textbox(label='Abstract', interactive=False, visible=False) ### RELEVANT PARTS (HIGHLIGHTS) with gr.Row(): with gr.Column(scale=2): # text from submission source_sentences = gr.Radio( choices=[], visible=False, label='Sentences from Submission Abstract', ) with gr.Column(scale=3): # highlighted text from paper highlight = gr.components.Interpretation(paper_abstract) ### EVENT LISTENERS # retrieve similar papers and show top results compute_btn.click( fn=get_similar_paper, inputs=[ abstract_text_input, pdf_file_input, author_id_input ], outputs=[ selected_papers_radio, source_sentences, paper_title1, # paper info affinity1, sent_pair_score1_1, sent_pair_source1_1, sent_pair_source1_1_hl, sent_pair_candidate1_1, sent_pair_candidate1_1_hl, sent_pair_score1_2, sent_pair_source1_2, sent_pair_source1_2_hl, sent_pair_candidate1_2, sent_pair_candidate1_2_hl, paper_title2, affinity2, sent_pair_score2_1, sent_pair_source2_1, sent_pair_source2_1_hl, sent_pair_candidate2_1, sent_pair_candidate2_1_hl, sent_pair_score2_2, sent_pair_source2_2, sent_pair_source2_2_hl, sent_pair_candidate2_2, sent_pair_candidate2_2_hl, paper_title3, affinity3, sent_pair_score3_1, sent_pair_source3_1, sent_pair_source3_1_hl, sent_pair_candidate3_1, sent_pair_candidate3_1_hl, sent_pair_score3_2, sent_pair_source3_2, sent_pair_source3_2_hl, sent_pair_candidate3_2, sent_pair_candidate3_2_hl, see_more_rel_btn ] ) # Get more info (move to more interactive portion) see_more_rel_btn.click( fn=show_more, inputs=None, outputs=[ selected_papers_radio, source_sentences, title_row, paper_abstract ] ) # change highlight based on selected sentences from submission source_sentences.change( fn=change_output_highlight, inputs=[ selected_papers_radio, source_sentences ], outputs=highlight ) # change paper to show based on selected papers selected_papers_radio.change( fn=change_paper, inputs=selected_papers_radio, outputs= [ paper_title, paper_abstract, affinity, highlight ] ) gr.Markdown( """ --------- **Disclaimer.** This tool and its output should not serve as the sole justification for confirming a match for the submission. It is intended as a supplementary tool that the user may use at their discretion; the correctness of the output of the tool is not guaranteed. This may be improved by updating the internal models used to compute the affinity scores and sentence relevance, which may require additional research independently. The tool does not compromise the privacy of the reviewers as it relies only on their publicly-available information (e.g., names and list of previously published papers). """ ) if __name__ == "__main__": demo.launch()