GithubHelp home page GithubHelp logo

fastforwardlabs / summarize. Goto Github PK

View Code? Open in Web Editor NEW
15.0 15.0 6.0 87.07 MB

Summarize. is a Streamlit application that performs automatic text summarization using both extractive and abstractive models.

License: Apache License 2.0

Python 100.00%

summarize.'s People

Contributors

gaffney2010 avatar melaniebeck avatar victordibia avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

summarize.'s Issues

Highlighter takes too long for long text/summary

This example is used in the app and highlight doesn't work fast enough to be reasonable for the user experience.

Input text:

Each game of Jeopardy! features three contestants competing in three rounds: Jeopardy!, Double Jeopardy!, and Final Jeopardy! Each round involves trivia clues phrased as answers, to which the contestants must respond in the form of a question. For example, if a contestant were to select "Presidents for $200", the resulting clue could be "This 'Father of Our Country' didn't really chop down a cherry tree", to which the correct response is "Who is/was George Washington?"

The Jeopardy! and Double Jeopardy! rounds each have their own game boards. These boards consist of six categories with five clues each; the clues are concealed by dollar amounts from lowest to highest, ostensibly by difficulty. The values of the clues increased over time, with those in the Double Jeopardy! round always being double the range of the Jeopardy! round. On the original Jeopardy! series, clue values in the first round ranged from $10 to $50 in the Jeopardy! round and $20 to $100 in Double Jeopardy! On The All-New Jeopardy!, they ranged from $25 to $125 and $50 to $250. The 1984 series' first-round originally ranged from $100 to $500 in Jeopardy! and $200 to $1,000 in Double Jeopardy! these ranges were increased to $200-$1,000 and $400-$2,000, respectively, on November 26, 2001.Gameplay begins when the returning champion selects any position on the game board; if there is no returning champion, contestant order is chosen via a random draw prior to the show. The underlying clue is revealed and read aloud by the host, after which any contestant may ring in using a lock-out device. The first contestant to ring in successfully is prompted to respond to the clue by stating a question containing the correct answer to the clue; any grammatically coherent question with the correct answer within it counts as a correct response. If the contestant responds correctly, its dollar value is added to the contestant's score, and they may select a new clue from the board. An incorrect response, or a failure to respond within five seconds, deducts the clue's value from the contestant's score and allows the other contestants the opportunity to ring in and respond. If the response is not technically incorrect but otherwise judged too vague to be correct, the contestant is given additional time to provide a more specific response. Whenever none of the contestants ring in and respond correctly, the host gives the correct response, and the player who selected the previous clue chooses the next clue. Gameplay continues until the board is cleared or the round's time length expires, which is typically indicated by a buzzer sound. The contestant who has the lowest score selects the first clue to start the Double Jeopardy! round.A "Daily Double" is hidden behind one clue in the Jeopardy! round, and two in Double Jeopardy! The name and inspiration were taken from a horse-racing term. Daily Double clues with a sound or video component are known as "Audio Daily Doubles" and "Video Daily Doubles" respectively. Before the clue is revealed, the contestant who has selected the Daily Double must declare a wager, from a minimum of $5 to a maximum of his or her entire score (known as a "true Daily Double") or the highest clue value available in the round, whichever is greater. Only the contestant who chooses the Daily Double is allowed to answer the clue and they must provide an answer. A correct response adds the value of the wager to the contestant's score while an incorrect response (or failure to provide any response at all) deducts the same value. Whether or not the contestant responds correctly, they choose the next clue.During the Jeopardy! round, except in response to the Daily Double clue, contestants are not penalized for forgetting to phrase their response in the form of a question, although the host reminds contestants to watch their phrasing in future responses. In the Double Jeopardy! round and in the Daily Double in the Jeopardy! round, the phrasing rule is followed more strictly, with a response not phrased in the form of a question counting as wrong if it is not re-phrased before the host or judges make a ruling.Contestants are encouraged to select the clues in order from lowest to highest value, as the clues are written in each category to flow from one to the next, as is the case with game shows that ask questions in a linear string. Deviating from this is known as the "Forrest Bounce", a strategy in which contestants randomly pick clues to confuse opponents that was first used in 1985 by Chuck Forrest, who won over $70,000 in his initial run as champion. According to Trebek, this strategy not only annoyed him but the staffers as well since it also disrupts the rhythm that develops when revealing the clues and increases the potential for error. Another strategy used by some contestants is to play all of the higher-valued clues first and build up a substantial lead. James Holzhauer, whose April–June 2019 winning streak included the ten highest single-day game totals, regularly used this strategy, in conjunction with the Forrest Bounce and aggressive Daily Double wagering.From the premiere of the original Jeopardy! until the end of the 1984–85 syndicated season, contestants were allowed to ring in as soon as the clue was revealed. Since September 1985, contestants are required to wait until the clue is read before ringing in. To accommodate the rule change, lights were added to the game board (unseen by home viewers) to signify when it is permissible for contestants to signal; attempting to signal before the light goes on locks the contestant out for half of a second. The change was made to allow the home audience to play along more easily and to keep an extremely fast contestant from potentially dominating the game. In pre-1985 episodes, a buzzer sounded when a contestant signaled; according to Trebek, the buzzer was eliminated because it was "distracting to the viewers" and sometimes presented a problem when contestants rang in while Trebek was still reading the clue. Contestants who are visually impaired or blind are given a card with the category names printed in Braille before each round begins, and an audible tone is played after the clue has been read aloud.
If it is determined that a previous response was wrongly ruled correct or incorrect during the taping of an episode, the scores are adjusted at the first available opportunity. If an error that may have affected the result is not discovered until after taping of an episode is completed, the affected contestant(s) are invited back to compete on a future show, complying with federal quiz show regulations.Contestants who finish Double Jeopardy! with $0 or a negative score are automatically eliminated from the game at that point and awarded the third-place prize. On at least one episode hosted by Art Fleming, all three contestants finished Double Jeopardy! with $0 or less, and as a result, no Final Jeopardy! round was played. This rule is still in place for the syndicated version, although staff has suggested that it is not set in stone and they may decide to display the clue for home viewers' play if such a situation were ever to occur.

Summary:

 Each game of Jeopardy! features three contestants competing in three rounds. Each round involves trivia clues phrased as answers, to which the contestants must respond in the form of a question. For example, if a contestant were to select "Presidents for $200", the resulting clue could be "This \'Father of Our Country\' didn\'t really chop down a cherry tree"\n Gameplay begins when the returning champion selects any position on the game board. If there is no returning champion, contestant order is chosen via a random draw prior to the show. The underlying clue is revealed and read aloud by the host, after which any contestant may ring in using a lock-out device. The first contestant to ring in successfully is prompted to respond to the clue by stating a question containing the correct answer. The contestant who has the lowest score selects the first clue to start the Double Jeopardy! round.\n Only the contestant who chooses the Daily Double is allowed to answer the clue and they must provide an answer. A correct response adds the value of the wager to the contestant\'s score while an incorrect response (or failure to provide any response at all) deducts the same value. Since September 1985, contestants are required to wait until the clue is read before ringing in. Lights were added to the game board to signify when it is permissible for contestants to signal.\n If it is determined that a previous response was wrongly ruled correct or incorrect during the taping of an episode, the scores are adjusted at the first available opportunity. If an error that may have affected the result is not discovered until after taping, the affected contestant(s) are invited back to compete on a future show.

AMP Review 1

Functional Review:

  • In Step 2: Install Dependencies, the spaCy model download fails silently because the command uses !python instead of !python3. This consequently results in a UI error in the Streamlit app. I forked the project and tested this simple change (python3 instead of python) and it works as expected - so should be a quick fix for you to commit.
    image

image

Content Review:

Streamlit App

  • Placeholder Task - noting this here so we don't forget to populate the blog post links once available (NOTE there may be more than just these):
    image
  • "Qualitative Comparison" --> "Quantitative Comparison": When a user switches from the Wikipedia dataset to the Daily News, the title of the "comparison" card should change to reflect that we now have the ability to quantify the produced summary, right?
  • UX Suggestion: Its slightly confusing to me as a first time user that the language here says: "Choose a passage to summarize." because the subsequent dropdown isn't actually passages, it's data sources.... Maybe a better instruction could be "Select a dataset" or "Select a data source to work with"
    image

README

  • Suggestion: The first sentence of the README makes reference to the legacy "Brief" application. Is there anything we can link to for this? I don't think the prototype is still up but maybe a blogpost or report something that helps people understand what this was?
  • Edit: I think we can remove the "fiction book chapters" since that was originally intended but didn't work out as well as expected...right?
    image
  • Placeholder Task: reminder to update this link once the url exists
    image
  • Edit: This should probably be a python3 command, right?
    image
  • Edit: I think the card in the Prototype Catalog will now say "Summarize." right?
    image

Misc

  • Looks like we need to add the Apache license file to the repo
  • For the "references to borrowed code", it looks like one of them includes an inline comment in the code with the link to huggingfaces source. The second one looks to be missing that comment + link. Can you add that so both are referenced inline? Thanks!
  • Can you complete/confirm the content in columns C-F for Default Catalog metadata in the tracking doc
  • Can you drop a thumbnail image into the review doc that will appear in CML AMP Catalog?

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.