Summarize any Article as a Dialog Between two Individuals with Bing Chat
I lately found Recast, a superb device that turns your want-to-read articles into wealthy audio summaries. It summarises lengthy articles into the human dialog.
Right here is an instance of Recast audio file:
As a result of I used to be fascinated by this concept I attempted to duplicate it with the assistance of ChatGPT Bing Chat. My thought is straightforward, use Bing Chat to create a dialog abstract and later use Microsoft Edge Browser for text-to-speech conversion.
Right here is the immediate for Bing Chat:
browse https://electrek.co/2023/05/15/renewables-us-utility-scale-generating-capacity-q1-2023/ and make a abstract within the type of a dialog between two folks. Go away an empty row after every dialog. Bolden the vital elements of the textual content. Earlier than the dialog begin with 2 or 3 sentence abstract.Change the URL along with your URL.
And the Bing Chat Response:
So you could have your abstract in dialog type.
It additionally works with PDFs.
browse https://arxiv.org/pdf/2305.07185.pdf and make a abstract in a type of dialog between two folks. Go away an empty row after every dialog. Bolden the vital elements of the textual content. Earlier than the dialog begin with 2 or 3 sentence abstract.And Bing Chat response:
The paper proposes Megabyte, a multi-scale decoder structure that allows end-to-end differentiable modeling of sequences of over a million bytes. Megabyte segments sequences into patches and makes use of a neighborhood submodel inside patches and a worldwide mannequin between patches. The paper exhibits that Megabyte can carry out competitively with subword fashions on lengthy context language modeling, obtain state-of-the-art density estimation on ImageNet, and mannequin audio from uncooked information.
A attainable dialog primarily based on the paper is:
**Alice**: Hey Bob, have you ever learn the brand new paper on Megabyte?
**Bob**: No, what's that?
**Alice**: It is a new mannequin for predicting million-byte sequences with multiscale transformers.
**Bob**: Million-byte sequences? That sounds difficult.
**Alice**: Yeah, it's. Most current autoregressive transformers can solely deal with brief sequences like textual content or low-resolution photographs. They scale poorly to lengthy sequences like high-resolution photographs, podcasts, code, or books.
**Bob**: So how does Megabyte resolve this downside?
**Alice**: Nicely, it makes use of a intelligent trick. It segments the sequences into patches and makes use of a neighborhood submodel inside patches and a worldwide mannequin between patches.
**Bob**: What does that imply?
**Alice**: It implies that it might probably use sub-quadratic self-attention, a lot bigger feedforward layers, and improved parallelism throughout decoding. This makes it quicker and extra environment friendly than normal transformers.
**Bob**: I see. And does it work effectively?
**Alice**: Sure, it does. The paper exhibits that Megabyte can carry out competitively with subword fashions on lengthy context language modeling, obtain state-of-the-art density estimation on ImageNet, and mannequin audio from uncooked information.
**Bob**: Wow, that is spectacular. I ponder what sort of purposes it might probably allow.
**Alice**: Me too. Perhaps we are able to use it to generate high-quality content material with out tokenization.
**Bob**: That might be superior. I will take a look at the paper. Thanks for telling me about it.
Supply: Dialog with Bing, 17/5/2023
(1) [2305.07185] MEGABYTE: Predicting Million-byte Sequences with .... https://arxiv.org/abs/2305.07185.
(2) arXiv.org. https://arxiv.org/pdf/2205.07185.pdf.
(3) PDF for 2305.04185 - arXiv.org. https://arxiv.org/pdf/2305.04185.However why not make it one step additional and make Microsoft Edge learn the textual content?
The issue with Microsoft Edge is that text-to-speech all the time begins from the highest of the webpage. However we solely what to learn the principle textual content generated with Bing Chat. I remembered that again within the day I made the Writability – Minimal Distractions Free Browser Text Writing Tool that may assist us copy the textual content within the browser and pressure the Edge to learn it.
You must paste this code contained in the Microsft Edge URL
information:textual content/html;charset=utf-8, <title>Writability</title><physique OnLoad='doc.physique.focus();' contenteditable type="font-size:21px;line-height:1.6;font-family:'Chaparral Professional',Georgia;max-width:21em;margin:0 auto;padding:3rem;background-color:rgb(233,233,225);shade:rgb(68,68,68);" spellcheck="false">And paste any textual content for Edge text-to-speech to learn it.
It seems one thing like this:
The ultimate end result seems like this:
So there you could have it your individual non-public dialog Summarizer and text-to-voice reader.