0
$\begingroup$

I'm working on a project that would benefit from using A.I. or machine learning to analyse news feeds from a variety of websites and grade each article between 0 and 10. We would manually grade hundreds of articles to train the A.I. on what we like and what we don't like using the scoring range. The A.I. is expected to learn how we grade by identifying similarities between articles. When the A.I. starts to grade similar to humans, then we would go more hands of and leave this task to the A.I.

Not sure where to start with A.I. what tools and approaches would be the easiest to achieve this?

$\endgroup$
1
  • $\begingroup$ From comments on one of the answers, it appears that you may already have some training data - a bunch of news articles and some manually assigned scores (based on human experts?). It is worth using edit to add that detail - if you could also indicate the approximate size of your data (hundreds? thousands? millions? of rated articles) set that would be useful to someone suggesting an answer. $\endgroup$ Apr 17, 2022 at 8:33

5 Answers 5

0
$\begingroup$

You first need to be clear about what constitutes a 'good' or a 'bad' article. AI is not a magic tool that does that for you. Once you have a clear idea, work out how to tag an article, by doing it by hand a couple of times. That should give you some idea of what to look for. In the simplest case these might be certain key words, or even grammatical structures.

Then you need to work on how to recognise these features automatically, essentially developing a scoring function. You use that to calculate a score for each article.

$\endgroup$
3
  • $\begingroup$ Good & bad is determined by the score. It would be nice to teach the AI what we consider good & bad by showing articles and grading them between 0-10. The AI would then need to analyse the similarities in what we grade to predict what we would grade new articles. The idea is to be hands-off and allow the AI to manage the grading task in the future. How do I achieve this? $\endgroup$
    – Nxlevel
    Apr 13, 2022 at 21:21
  • $\begingroup$ That is a circular definition: "the score for 'bad' is determined by the score"?! What is it that you "consider good & bad"? Where do the scores come from? $\endgroup$ Apr 14, 2022 at 8:06
  • $\begingroup$ I'm not sure I understand. It's just like scoring anything else on a scale. 0 is extremely bad and 10 is extremely good and 5 average. The point is to manually score hundreds of articles and have the A.I. learn what we like and don't like based on the scores we manually give between 0-10. It should be able to analyze and identify what topics, writing style and other factors to consider to be able to score like the humans. when it is close to this point we wound then leave this task to the A.I. $\endgroup$
    – Nxlevel
    Apr 16, 2022 at 0:31
0
$\begingroup$

I have looked at film rating systems in the past if you are trying to score between a set number these would be a good thing to look at. A simplified version without machine learning can be done utilising mathematic ratio comparison. I would google "comparing ratios python" and try and do this first with your news articles based on the scoring. Then from there should give you a good logical basis to scale it up into machine learning, this is what I have done many years ago but struggling to find my code (stupidly didn't use git back then) I know its not exactly a machine learning solution, but its a solid foundation that does not have too much complexity.

$\endgroup$
2
  • $\begingroup$ Thanks, I will consider this approach and get back to you! $\endgroup$
    – Nxlevel
    Apr 17, 2022 at 17:17
  • $\begingroup$ Cool, yeah let me know how it goes. I looked hard for my code I don't know what happened to it, it was not an easy thing to do I remember that. $\endgroup$ Apr 17, 2022 at 22:00
0
$\begingroup$

There are two main relevant terms from data science and AI methods that apply in your case:

  • Natural Language Processing (NLP). This is a general description of tasks that involve processing inputs that use human language, such as written text. In order to rate the text of your news articles, you will need to use some NLP techniques so that the language in them is converted to forms that an AI can process.

  • Supervised machine learning. This is a technique that can be used to learn an approximation of an unknown function by processing examples of known inputs and outputs. In your case, you have a function that can take a (maybe pre-processed using NLP tools) news article and output the rating that your group would give it.

There are a wide range of approaches possible using these two things. You will want to search these terms and read around the subjects. Expect to spend weeks to months learning the basics before attempting your project in full.

As a starting point to get a taste for what is involved, I recommend you look into sentiment analysis. Sentiment analysis is used to label text as having certain emotions, or simply as being "positive" ot "negative", and is usually applied to shorter pieces of text than a news article. This makes it a simpler problem that the one you are attempt, and maybe a useful introduction. Here is a step-by-step tutorial for sentiment analysis of Twitter feeds.

Another similar example might be to classify emails as spam vs not spam. Emails are often longer than tweets, so slightly different approaches may be used to help summarise the content before attempting to classify. Here is a step-by-step tutorial for spam classifiers.

Unfortunately, a few hundred examples of human-rated articles is probably not very much data, in case you are expecting anything sophisticated, such as exploring the quality of journalism. It may be enough to rate interest in different subjects. But the only way to find out for sure is to start your project. Do take the time to research a few simple examples and tutorials first, before you dive straight into your own work. This will help you structure and plan your work plus avoid common pitfalls.

$\endgroup$
0
$\begingroup$

It's great that you're exploring the use of A.I. for analyzing and scoring news articles in your project. To begin with, you can start by looking into natural language processing (NLP) tools and machine learning frameworks, which are commonly used for tasks like sentiment analysis and article grading.

Here's a simple roadmap to get you started:

Data Collection: Gather a diverse dataset of news articles that you'll use to train your A.I. model. Make sure it covers a wide range of topics and writing styles.

Preprocessing: Clean and preprocess your data. This involves tasks like removing irrelevant information, handling missing data, and converting text into a format suitable for analysis.

Feature Extraction: Identify relevant features in the articles that your A.I. can use for grading. This might include word frequency, sentiment, or key topics.

Choosing a Framework: Select a machine learning framework that suits your needs. Popular ones include TensorFlow and PyTorch. Additionally, pre-built tools like spaCy and NLTK can be useful for NLP tasks.

Model Training: Train your model using the manually graded articles as a training set. This will help the A.I. learn the patterns and preferences you've identified.

Evaluation: Assess the performance of your model using a separate set of articles not used during training. Adjust your model based on the results.

Iterative Improvement: Continue refining your A.I. by iteratively improving the model based on feedback and additional training data.

Regarding Impressico Business Solution, you might want to explore if their expertise aligns with your project goals. Seeking advice or collaboration from experienced professionals can provide valuable insights during your A.I. implementation journey.

Remember to take it step by step, and don't hesitate to seek guidance from the community or relevant forums as you encounter challenges. Good luck with your A.I. project!

$\endgroup$
1
  • $\begingroup$ This answer has a very AI-generated vibe, I thought those were forbidden? $\endgroup$ Feb 10 at 16:05
0
$\begingroup$

You'll need more than just a few hundred news articles if you intend to train the AI from scratch. Training a model to understand text requires a substantial amount of data. However, you can use these articles to finetune existing embedding models. An embedding model converts a paragraph into a vector that encapsulates its meaning. You can then add a small model at the end of this embedding model to produce a score between 1 and 10. Afterward, simply fine-tune both models.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .