Boosting Elasticsearch with machine learning – Elasticsearch, RankLib, Docker

Telescope

Elastic search is powerful search engine. Its distributed architecture give ability to build scalable full-text search solution. Additionally it provides comprehensive query language.

Despite this sometimes the engine and search results is not enough to meet the expectations of users. In such situations it is possible to boost search quality using machine learning algorithms.

In this article I will show how to do this using RankLib library and LambdaMart algorithm . Moreover I have created ready to use platform which:

  1. Index the data
  2. Helps to label the search results in the user friendly way
  3. Trains the model
  4. Deploys the model to elastic search
  5. Helps to test the model

The whole project is setup on the docker using docker compose thus you can setup it very easy.
The platform is based on the elasticsearch learning to rank plugin. I have also used the python example described in this project.

Before you will start you will need docker and docker-compose installed on your machine (https://docs.docker.com/get-started/)

To run the project you have to clone it:

git clone https://github.com/qooba/elasticsearch-learning-to-rank.git

Then to make elasticsearch working you need to create data folder with appropriate access:

cd elasticsearch-learning-to-rank/
mkdir docker/elasticsearch/esdata1
chmod g+rwx docker/elasticsearch/esdata1
chgrp 1000 docker/elasticsearch/esdata1

Finally you can run the project:

docker-compose -f app/docker-compose.yml up

Now you can open the http://localhost:8020/.

1. Architecture

There are three main components:

A. The ngnix reverse proxy with angular app
B. The flask python app which orchestrates the whole ML solution
C. The elastic search with rank lib plugin installed

A. Ngnix

I have used the Ngnix reverse proxy to expose the flask api and the angular gui which helps with going through the whole proces.

ngnix.config

server {
    listen 80;
    server_name localhost;
    root /www/data;

    location / {
        autoindex on;
    }

    location /images/ {
        autoindex on;
    }

    location /js/ {
        autoindex on;
    }

    location /css/ {
        autoindex on;
    }

    location /training/ {
        proxy_set_header   Host                 $host;
        proxy_set_header   X-Real-IP            $remote_addr;
        proxy_set_header   X-Forwarded-For      $proxy_add_x_forwarded_for;
        proxy_set_header   X-Forwarded-Proto    $scheme;
        proxy_set_header Host $http_host;

        proxy_pass http://training-app:5090;
    }
}

B. Flask python app

This is the core of the project. It exposes api for:

  • Indexing
  • Labeling
  • Training
  • Testing

It calls directly the elastic search to get the data and do the modifications.
Because training with RankLib require the java thus Docker file for this part contains default-jre installation. Additionally it downloads the RankLib-2.8.jar and tmdb.json (which is used as a default data source) from: http://es-learn-to-rank.labs.o19s.com/.

Dockerfile

FROM python:3

RUN \
    apt update && \
    apt-get -yq install default-jre
RUN mkdir -p /opt/services/flaskapp/src
COPY . /opt/services/flaskapp/src
WORKDIR /opt/services/flaskapp/src
RUN pip install -r requirements.txt
RUN python /opt/services/flaskapp/src/prepare.py
EXPOSE 5090
CMD ["python", "-u", "app.py"]

C. Elastic search

As mentioned before it is the instance of elastic search with the rank lib plugin installed

Dockerfile

FROM docker.elastic.co/elasticsearch/elasticsearch:6.2.4
RUN /usr/share/elasticsearch/bin/elasticsearch-plugin install \ 
-b http://es-learn-to-rank.labs.o19s.com/ltr-1.1.0-es6.2.4.zip

All layers are composed with docker-compose.yml:

version: '2.2'
services:
  elasticsearch:
    build: ../docker/elasticsearch
    container_name: elasticsearch
    environment:
      - discovery.type=single-node
      - bootstrap.memory_lock=true
      - xpack.security.enabled=false
      - "ES_JAVA_OPTS=-Xms512m -Xmx512m"
    ulimits:
      memlock:
        soft: -1
        hard: -1
    volumes:
      - ../docker/elasticsearch/esdata1:/usr/share/elasticsearch/data
    networks:
      - esnet

  training-app:
    build: ../docker/training-app
    networks:
      - esnet
    depends_on:
      - elasticsearch
    environment:
      - ES_HOST=http://elasticsearch:9200
      - ES_INDEX=tmdb
      - ES_TYPE=movie
    volumes:
      - ../docker/training-app:/opt/services/flaskapp/src

  nginx:
    image: "nginx:1.13.5"
    ports:
      - "8020:80"
    volumes:
      - ../docker/frontend-reverse-proxy/conf:/etc/nginx/conf.d
      - ../docker/frontend-reverse-proxy/www/data:/www/data
    depends_on:
      - elasticsearch
      - training-app
    networks:
      - esnet


volumes:
  esdata1:
    driver: local

networks:
  esnet:

2. Platform

The platform helps to run and understand the whole process thought four steps:

A. Indexing the data
B. Labeling the search results
C. Training the model
D. Testing trained model

A. Indexing

The first step is obvious thus I will summarize it shortly. As mentioned before the default data source is taken from tmdb.json file but it can be simply changed using ES_DATA environment variable in the docker-compose.yml :

training-app:
    environment:
      - ES_HOST=http://elasticsearch:9200
      - ES_DATA=/opt/services/flaskapp/tmdb.json
      - ES_INDEX=tmdb
      - ES_TYPE=movie
      - ES_FEATURE_SET_NAME=movie_features
      - ES_MODEL_NAME=test_6
      - ES_MODEL_TYPE=6
      - ES_METRIC_TYPE=ERR@10

Clicking Prepare Index the data is taken from ES_DATA file and indexed in the elastic search.

prepare index

Additionally you can define:
ES_HOST – the elastic search url
ES_USER/ES_PASSWORD – elastic search credentials, by default authentication is turned off
ES_INDEX/ES_TYPE – index/type name for data from ES_DATA file
ES_FEATURE_SET_NAME – name of container for defined features (described later)
ES_MODEL_NAME – name of trained model kept in elastic search (described later)
ES_MODEL_TYPE – algorithm used to train the model (described later).
ES_METRIC_TYPE – metric type (described later)

We can train and keep multiple models in elastic search which can be used for A/B testing.

B. Labeling

The supervised learning algorithms like learn to rank needs labeled data thus in this step I will focus on this area.
First of all I have to prepare the file label_list.json which contains the list of queries to label e.g.:

[
    "rambo",
    "terminator",
    "babe",
    "die hard",
    "goonies"
]

When the file is ready I can go to the second tab (Step 2 Label).

label

For each query item the platform prepare the result candidates which have to be ranked from 0 to 4.

You have to go through the whole list and at the last step
the labeled movies are saved in the file :

# grade (0-4)   queryid docId   title
# 
# Add your keyword strings below, the feature script will 
# Use them to populate your query templates 
# 
# qid:1: rambo
# qid:2: terminator
# qid:3: babe
# qid:4: die hard
# 
# https://sourceforge.net/p/lemur/wiki/RankLib%20File%20Format/
# 
# 
4 qid:1 # 7555 Rambo
4 qid:1 # 1370 Rambo III
4 qid:1 # 1368 First Blood
4 qid:1 # 1369 Rambo: First Blood Part II
0 qid:1 # 31362 In the Line of Duty: The F.B.I. Murders
0 qid:1 # 13258 Son of Rambow
0 qid:1 # 61410 Spud
4 qid:2 # 218 The Terminator
4 qid:2 # 534 Terminator Salvation
4 qid:2 # 87101 Terminator Genisys
4 qid:2 # 61904 Lady Terminator
...

Each labeling cycle is saved to the separate file: timestamp_judgments.txt

C. Training

Now it is time to use labeled data to make elastic search much more smarter. To do this we have to indicate the candidates features.
The features list is defined in the files: 1-4.json in the training-app directory.
Each feature file is elastic search query eg. the {{keyword}}
(which is searched text) match the title property:

{
    "query": {
        "match": {
            "title": "{{keywords}}"
        }
    }
}

In this example I have used 4 features:
– title match keyword
– overview match keyword
– keyword is prefix of title
– keyword is prefix of overview

I can add more features without code modification, the list of features is defined and read using naming pattern (1-n.json).

Now I can go to the Step 3 Train tab and simply click the train button.

train

At the first stage the training app takes all feature files and build the features set which is save in the elastic search (the ES_FEATURE_SET_NAME environment variable defines the name of this set).

In the next step the latest labeling file (ordered by the timestamp) is processed (for each labeled item the feature values are loaded) eg.

4 qid:1 # 7555 Rambo

The app takes the document with id=7555 and gets the elastic search score for fetch defined feature.
The Rambo example is translated into:

4   qid:1   1:12.318446 2:10.573845 3:1.0   4:1.0 # 7555    rambo

Which means that score of feature one is 12.318446 (and respectively 10.573845, 1.0, 1.0 for features 2,3,4 ).
This format is readable for the RankLib library. And the training can be perfomed.
The full list of parameters is available on: [https://sourceforge.net/p/lemur/wiki/RankLib/][https://sourceforge.net/p/lemur/wiki/RankLib/].

The ranker type is chosen using ES_MODEL_TYPE parameter:
– 0: MART (gradient boosted regression tree)
– 1: RankNet
– 2: RankBoost
– 3: AdaRank
– 4: Coordinate Ascent
– 6: LambdaMART
– 7: ListNet
– 8: Random Forests

The default used value is LambdaMART.

Additionally setting ES_METRIC_TYPE we can use the optimization metric.
Possible values:
– MAP
– NDCG@k
– DCG@k
– P@k
– RR@k
– ERR@k

The default value is ERR@10

train

Finally we obtain the trained model which is deployed to the elastic search.
The project can deploy multiple trained models and the deployed model name is defined by ES_MODEL_NAME.

D. Testing

In the last step we can test trained and deployed model.

test

We can choose the model using the ES_MODEL_NAME parameter.

It is used in the search query and can be different in each request which is useful when we need to perform A/B testing.

Happy searching 🙂

Another brick in the … recommendation system – Databricks in action

Brick

Today I’d like to investigate the Databricks. I will show how it works and how to prepare simple recommendation system using collaborative filtering algorithm which can be used to help to match the product to the expectations and preferences of the user. Collaborative filtering algorithm is extremely useful when we know the relations (eg ratings) between products and the users but it is difficult to indicate the most significant features.

Databricks

First of all, I have to setup the databricks service where I can use Microsoft Azure Databricks or Databricks on AWS but the best way to start is to use the Community version.

Data

In this example I use the movielens small dataset to create recommendations for the movies. After unzipping the package I use the ratings.csv file.
On the main page of Databrick click Upload Data and put the file.

The file will be located on the DBFS (Databrick file system) and will have a path /FileStore/tables/ratings.csv. Now I can start model training.

Notebook

The data is ready thus in the next step I can create the databricks notebook (new notebook option on the databricks main page) similar to Jupyter notebook.

Using databricks I can prepare the recommendation in a few steps:

First of all I read and parse the data, because the data file contains the header additionally I have to cut it. In the next step I split the data into training which will be used to train the model and testing part for model evaluation.

I can simply create the ratings for each user/product pair but also export user and products (in this case movies) features. The features in general are meaningless factors but deeper analysis and intuition can give them meaning eg movie genre. The number of features is defined by the rank parameter in training method (used for model training).
The user/product rating is defined as a scalar product of user and product feature vectors.
This gives us ability to use them outside the databricks eg in relational database prefilter the movies using defined business rules and then order using user/product features.

Finally I have shown how to save the user and product features as a json and put it to Azure blob.

Hello from serverless messanger chatbot

Chatbot

Messanger chatbots are now becoming more and more popular. They can help us order pizzas, ask about the weather or check the news.

In this article, I would like to show you how to build a simple messanger chatbot in python and pass it on AWS lambda. Additionally use the wit.ai service to add to it the natural language understanding functionality and make it more intelligent.

To build the messanger chatbot I will need facebook app and facebook page.

Facebook page

The whole communication is going through a Facebook page thus I need to create it
https://www.facebook.com/bookmarks/pages

I will need the page id which you can find at the bottom of your page:
https://www.facebook.com/page_name/about

Facebook app

Then I create facebook app.
https://developers.facebook.com/quickstarts/?platform=web

Settings

I will copy the AppId and AppSecret which will be needed in the next steps:

Messanger product

Then I will add the messanger product and setup it.
I need to select page we already created and copy generated access token.

 

Webhook

Finally I have to setup the webhooks for messanger product
To finish this step I need to setup our chatbot on aws lambda.
I also have to provide the verify token which will be used to validate our endpoint.

AWS Lambda

Now I will prepare my chatbot endpoint. I will setup it on the AWS lambda.

Trigger

For my chatbot I need to configure API Gateway.
I have to choose security open otherwise I won’t be able to call it from messanger

 

Code

I also need to provide code which will handle the messanger webhook and send response.
I will simply put the code in the online editor.
Let’s take a look at the code:

Configuration

Bellow I have to setup environment variables:
verify_token – verification token (I use keepass to generate it) which we will use in webhook setup
access_token – value from the messanger webhook page setup
app_secret – facebook app secret

Now I’m ready to finish the webhook configuration:

I use api gateway url ass Callback URL and verify_token I have just generated.

Natural language undestanding

Messanger give easy way to add natural language undestanding functionality. To add this I simply configure it on messanger product setup page

Here I can choose already trained models but I will go further and I will create custom model.
Messanger will create the new wit.ai project for me.

On the wit.ai I can simply add some intents (like: hungry) and additional information which can be retrieved from the phrase (like: I want some pizza)

The messanger/wit integration is very smooth let’s analyze the webhook json I get when I put I want to eat pizza

After wit integration the nlp object was added. Now I can get the recognized intent with some confidence (like: hungry) and additional entities (like: dish).

Finally I can talk with my chatbot 🙂