I tried HR Tech to develop an expert search engine by machine learning in-house meeting information

TL; DR

We will create an expert search engine by utilizing the dark data that sleeps in the company and the data related to meetings. Such a guy membersearch-min.png

This will allow you to find an in-house expert and a contact path to that expert.

1. Background. .. ..

1.1 What is work style reform?

Jitahara is often said to have work style reforms.

Jitahara: Forcing employees to leave the company, such as "Don't work overtime" or "Return on time", without specific measures to reduce overtime hours.

There is. Serious people will return as they were told ~~ You will have to take it home and work overtime. ** That's stupid! !! !! ** **

1.2 Meetings ... Would you like to reduce it?

According to Asahi Shimbun

Loss of wasteful meetings at large companies: 670,000 hours a year and 1.5 billion yen

That's right. I will listen to it straightforwardly.

** Can you reduce meetings? ** ** ** If we can reduce by 1.5 billion yen, shouldn't we be rewarded as much as Mr. Ghosn? ** **

The world is full of so much know-how! !! !! !!

kaigi_sakugen.png

But why doesn't it change over the years? .. .. .. .. Personally, after all, I don't think I'm serious about reducing the number of meetings and time. In short, the know-how that comes out from google is just a blunder.

1.3 Meeting ... Will you use the data?

Even if it is a venture, if the scale exceeds 300 people *, the things that each employee can do will be limited.

That doesn't mean you're not doing anything. Instead of reducing the number of meetings without thinking (abuse) as is done in your company or my company, let's take advantage of it ** even a little.

2. Issues and solution policies

2.1 Challenges that are likely to be a large company

As an issue that tends to occur in an organization where the number of employees including the group exceeds a certain number,

What's the matter? There is, right?

So, let's set this as an issue to be solved.

2.2 Solution policy

Isn't it accumulated in vain and is the first in Dark Data? I will try to utilize the conference data that seems to be.

Dark data Refers to "data that may create value within a company but is not utilized"

It is not used because it is only accumulated. Even if I googled, there were no examples of its use. What a waste.

2.3 Characteristics of meeting data

Now let's consider the characteristics of conference data.

merit

Demerit

There are some disadvantages, but ... let's try it first!

3. Data collection and preprocessing

For more information, please see here [https://boomin.yokohama/archives/1619)

4. Machine learning

It's been long, so see here.

5. Network analysis

It's been long, so see here.

6. Visualization

Vue + Vuex + [Vuetify](https://vuetifyjs.com/ja Developed using /).

6.1 First, the Home screen!

If the home screen is dull, some people may just go home. So, I did my best to ** feel cool on my own basis ** (misallocation)

hrsearch-min.png

6.2 Search for experts

Select Keyword Search from the left pane menu to search for people by word.

keywordsearch-min.png

To! If you search for the word machine learning, you will find ** the Minister of Education, Culture, Sports, Science and Technology **. By the way, AI looks like this.

AI-min.png

In addition, the color is a color for each cluster by clustering people who have similar remarks to some extent. It might have been better to set it by political party here, but it's ** annoying ** so I'll leave it as it is.

6.3 Search for connections

Next is the connection search. From Member's Search, search for Prime Minister Abe's connections (people with similar remarks here).

membersearch-min.png

I see. .. .. .. (When I looked at this in the parliamentary minutes data, I didn't get any impressions.) I'm sure it would be a little interesting if we could feed the actual meeting data in-house. .. .. ..

6.4 Search for connections

It's finally time to search for connections (what I wanted to do the most). Then Prime Minister Abe and his sword? Let's examine the connection with Finance Minister Taro Aso.

connection1-min.png

Based on the content of the remarks, Mr. Toshio Beppu is sandwiched on the way. This area depends on the trained model, so I'll leave it for the time being.

Next, let's examine the connection between Prime Minister Abe and Representative Taro Yamamoto. (Because he was a member of parliament who came up with something)

connection2-min.png

On the way, three people are caught. Is that so?

6.5 Suggestion of search target word

Since there are many words that can be searched and many people are targeted, validation is performed before issuing a search query. However, it is difficult to know what kind of word is OK, so when you start typing a word, we have implemented a function to display search word candidates.

support-min.png

Now you can search for any word, but you can rest assured!

7 Development impression

What I thought about making so far is that ** UX design and UI design are important **. .. .. ..

Preparation of learning data, cleansing, morphological analysis, machine learning, and API development were all not that difficult. I think it's because the goal was clear as it was.

However, I started to make the front end without having a clear image, so I repeated that it was neither uh nor uh. Thanks to that, it's hard to feel that the source code isn't cool either. .. .. ..

I don't think this screen is good for me, but I'm not sure what to do with it.

8 Finally, for work style reform

Here, I showed the screen based on the minutes of the Diet, but I decided to use the learning data for the meeting data of my company (without permission). I think that you will receive various opinions, so please take them positively.

That? Who knows this? Oh, I'm lucky to know the connection if I search with this YO!

Everyone thinks

Does this person do this kind of work? Then, let's talk a little and try to brainstorm the idea of the project!

I would be happy if you think about it.

Recommended Posts

I tried HR Tech to develop an expert search engine by machine learning in-house meeting information
[Machine learning] I tried to do something like passing an image
I tried to get an image by scraping
I tried to predict the presence or absence of snow by machine learning.
I tried to predict the change in snowfall for 2 years by machine learning
[Keras] I tried to solve a donut-type region classification problem by machine learning [Study]
I tried to move machine learning (ObjectDetection) with TouchDesigner
I tried to build an environment for machine learning with Python (Mac OS X)
I tried to compress the image using machine learning
I tried to verify the yin and yang classification of Hololive members by machine learning
I tried to implement anomaly detection by sparse structure learning
I tried machine learning to convert sentences into XX style
[Machine learning] I tried to summarize the theory of Adaboost
An introduction to machine learning
I was frustrated by Kaggle, so I tried to find a good rental property by scraping & machine learning
I tried to understand supervised learning of machine learning in an easy-to-understand manner even for server engineers 2
I tried to make an activity that collectively sets location information
I tried machine learning with liblinear
Machine learning meeting information for HRTech
Search for technical blogs by machine learning focusing on "easiness to understand"
(Machine learning) I tried to understand Bayesian linear regression carefully with implementation.
I tried to classify mnist numbers by unsupervised learning [PCA, t-SNE, k-means]
I tried to visualize the model with the low-code machine learning library "PyCaret"
I tried to classify Oba Hana and Emiri Otani by deep learning
I installed Python 3.5.1 to study machine learning
An introduction to OpenCV for machine learning
An introduction to Python for machine learning
I tried to organize the evaluation indexes used in machine learning (regression model)
I'm not an engineer at all, but I classified faces by machine learning
I tried to extract a line art from an image with Deep Learning
I tried to implement various methods for machine learning (prediction model) using scikit-learn.
I tried to process and transform the image and expand the data for machine learning
I tried to classify Oba Hana and Emiri Otani by deep learning (Part 2)