Overview

• 1: Getting Started
• 2: Command reference
• 3: Training file
• 4: Use cases
• 4.1: Anti-Money Laundry
• 4.2: Asset management
• 4.3: Accounting
• 4.4: Content classification
• 4.5: Talent matching
• 5: Prices

1 - Getting Started

Data classification is the process of organizing and categorizing data based on specific criteria or attributes. It helps make data more structured, accessible, and understandable.

Let’s see an example

Suppose you need to categorize a set of texts that contain mixed references to cats, actors, dogs, and other things that don’t matter.

You will need a CSV file containing a few examples. In this file, each example corresponds to a row that provides at least two fields:

• prototype: a text that exemplifies a typical element in the category specified in the second field
• class: a short text that represent a category name

These examples are used by EZClassifier to create a personalized model. You can add to the model as many prototypes and categories as you like, as long as there is at least one example for each category. EZClassifier works with any language, even when used simultaneously in the same text or the same data file and/or examples.

Once you created your model, you use it to classify your data stream. EZClassifier will add two additional fields to your data:

• class: that is the predicted category for the described row
• similarity: a number from 0 to 1 that represents the confidence of EZClassifier about its classification (1=maxumum confidence , 0=no confidence).

• model examples
• input data
• classified data

Try it out!

HW & SW Prerequisites

To run EZClassifier, you need any computer with Java 17+ installed. You can download the Java 17+ for your architecture (Linux, macOS, Windows) from the the official Java site.

Install EZClassifier

Next, download the latest version of the EZC JAR file to a directory of your choice.

Lastly, you’ll need an API Key to enable the services (see prices here). Proof of Concept (POC) and free plans are available upon request.

Create shortcut for launching the java command and set the TC_API_KEY environment with your api key:

For example, with bash, open a terminal and type:

export TC_API_KEY=HERE-IS-YOUR-API-KEY
alias ezc='java -jar ezc.jar' 

For example, in windows open a terminal windows (CMD) and type:

set TC_API_KEY=HERE-IS-YOUR-API-KEY
doskey ezc=java -jar "%USERPROFILE%\Downloads\ezc.jar" $*

Test that the system is working:

# be sure it reports a java version > 17 
ezc --version

Step 1: download example files

Here you can download some sample data:

• model source file: some examples about cat, dog, and actor descriptions
• data to be classified: some strings to classify

Step 2: create a model

Create a new model from your examples:

ezc model train --name=mymodel --header --input=examples.csv

Be sure to use as –input argument the path (relative or absolute) of the downloaded model source file

To list all the available models you can use ezc model ls

Step 3: classify your data using the created model:

ezc classify --name=mymodel --header --input=input-data.csv --output=result.csv --threshold=0.84

The result.csv file in your current directory will contain your classified data. Note that data with a confidence less than 0.84 are assigned to the “OTHER” category

cat input-data.csv | ezc classify --name=mymodel > result.csv

Remove your model with ezc model rm --name=mymodel.

2 - Command reference

All the services offered by EZClassifier can be accessed using the command through the ezc command. See the getting started section for the instructions for downloading and installing it.

General syntax

For java JRE 17+ is required

The command returns 0 on success or a value > 0 on failure. The logs are written on stdout by default and can be redirected.

All commands requires a valid api key in the environment variable API_KEY

export TC_API_KEY=<api-key>

Usage:

java -jar <path of the downloaded jar file> [-hV] [COMMAND]

A software agent that classifies text based on the provided prototypes

Options:

• -h, --help
  Show this help message and exit.

• -V, --version
  Print version information and exit.

Commands:

• model
  Manage models

• classify
  Perform classification

Command model train

Train the model

model train [-hHSV] -C=<labelIndex> [-e=<apiEndpoint>]
                    [-i=<inputFilename>] [-k=<apiKey>] -n=<name> -P=<textIndex>
                    [-W=<weightIndex>]

You can enhance an already trained model by adding new examples through multiple calls to the “model train” command.

Options:

• -C, --class-index=<labelIndex>
  The column index in the CSV file that contains the field with the class attached to the text (from 0). By default is 1

• -e, --endpoint=<apiEndpoint> Api endpoint. By default https://api.mopso.io/v1/tc

• -h, --help
  Show this help message and exit.

• -H, --header
  Will ignore the first line of the CSV file.

• -i, --input=<inputFilename>
  The file containing the training data, by default “-” that means std in (e.g. -i - ). The stream is supposed to be in CSV format and MUST contain two fields (“prototype” and “class”) and some additional fields.

• -k, --api-key=<apiKey>
  A registered API key. If not present, the value from the env variable TC_API_KEY is used.

• -n, --name=<name>
  Name of the model to train, create a new model if the name is not found.

• -P, --prototype-index=<textIndex>
  The column index in the CSV file that contains the field with the text to classify (from 0)

• -S, --strict
  Runs the program in strict mode: any partially recoverable exception thrown during the execution will stop the training process. If not run in strict mode, the application will try to compensate for as many errors as possible.

• -V, --version
  Print version information and exit.

• -W, --weight-index=<weightIndex>
  The column index in the CSV file that contains the field with the classification weight (from 0). Set to -1 if not present.

Command model ls

List models

model ls [-hV] [-e=<apiEndpoint>] [-k=<apiKey>]

Options:

• -e, --endpoint=<apiEndpoint> Api endpoint. By default https://api.mopso.io/v1/tc

• -h, --help
  Show this help message and exit.

• -k, --api-key=<apiKey>
  A registered API key. If not present, the value from the env variable TC_API_KEY is used.

• -V, --version
  Print version information and exit.

Command model rm

model rm [-hV] [-e=<apiEndpoint>] [-k=<apiKey>] -n=<name>

Models that are not used for more than 3 month are automatically deleted.

Options:

• -e, --endpoint=<apiEndpoint> Api endpoint. By default https://api.mopso.io/v1/tc

• -h, --help
  Show this help message and exit.

• -k, --api-key=<apiKey>
  A registered API key. If not present, the value from the env variable TC_API_KEY is used.

• -n, --name=<name>
  Name of the model to remove.

• -V, --version
  Print version information and exit.

Command classify

Usage:
classify [-hHSV] [-e=<apiEndpoint>] [-i=<inFilename>] [-I=<textIndex>] [-k=<apiKey>] -n=<name> [-o=<outFilename>] [-t=<threshold>] [-T=<threads>]

Perform classification

• -e, --endpoint=<apiEndpoint> Api endpoint. By default https://api.mopso.io/v1/tc

• -h, --help
  Show this help message and exit.

• -H, --header
  If the flag is present, the first line of the input file is copied to the output with added columns ‘CLASS’ and ‘SIMILARITY_SCORE’. By default, it is assumed that the input has no header.

• -i, --input=<inFilename>
  The input filename; “-” means stdin (e.g. -i - ). The file must be in CSV format.

• -I, --index=<textIndex>
  The index in the CSV file that contains the field to classify (from 0). By default, the index is 0.

• -k, --api-key=<apiKey>
  A registered API key. If not present, the value from the env variable TC_API_KEY is used.

• -n, --name=<name>
  Model name.

• --no-buffer Execute the program in interactive mode. Will ignore –input, –output and –header options.

• -o, --output=<outFilename>
  The output filename; “-” means stdout (e.g. -o - ).

• -S, --strict
  Runs the program in strict mode: any partially recoverable exception thrown during the execution (i.e. a classification that fails or a row that can’t be parsed) will stop the program, truncating the output to the last stable state. If not run in strict mode, the application will try to compensate for as many errors as it’s possible.

• -t, --threshold=<threshold>
  Number between 0 (not included) and 1 (included) that is used to determine whether a match ‘SIMILARITY_SCORE’ is too low to be considered valid. In this case, the ‘CLASS’ is set to ‘OTHER’. Default value is 0.84.

• -T, --threads=<threads>
  The number of parallel jobs to be used by the classification services, by default is 1. If more than 1 is used, the output order is not preserved. The value is capped to the number of CPU cores.

• -V, --version
  Print version information and exit.

3 - Training file

To train your model you need a CSV file containing an example for each row.

Each example can contain a set of fields, some mandatory some optional. The index, staring from 0, of the fields in the record can be communicated to EZClassifier model train command with a set of command options

If the index of the fileld doe not exists in the record or the field format is not a valid value, the default value is used:

If an header raw is present, please do not forget to use the --header option in model train command

Here is an example of a model trainig file to find strings about cats:

prototype, class, weight, bias
"Persian Cat: Known for their long, luxurious fur and sweet temperament, Persian cats are one of the most popular breeds", cat
"siamese animal", cat, 0.8
"from siam", cat, 0.8, -0.5

In this example:

• the first row is the header (to be discarded using -H option during model training)
• the second row assumes a weight = 1.0 and a bias = 0.0
• the third row assumes a weight = 8.8 and a bias = 0.0
• the fourth row assumes a weight = 8.8 and a bias = -0.5

4 - Use cases

Here are some use cases where EZClassifier can increase your productivity.

4.1 - Anti-Money Laundry

Mopso, a leader in AML software, uses their special models to detect suspicious transactions.

4.2 - Asset management

Use your customized models to classify assets in a room, in a building, in a finacial portfolio or in a whole city.

4.3 - Accounting

Automatically assign a reference in the chart of accounts to each entry of the general Ledger entries.

4.4 - Content classification

Are you tired of manually sorting through endless texts, ads, and web content? We have the perfect solution for you! Our cutting-edge system can classify texts, ads, posts, and entire web pages effortlessly, saving you time and effort.

🔍 What can it do?

• Buying/Selling: Easily distinguish between content meant for buying and selling.
• Type of Item: Identify the type of item being sold, whether it’s electronics, an house, clothing, or something else.
• Size Categories: Classify items based on their size, be it large, medium, small, or any other dimension.

🌟 Why Choose Us?

• Easy to customize: create you models just with few examples for each category and let our A.I. engine to do the rest.
• Accuracy: Our system ensures accurate classification, making sure each item is placed in the correct category.
• Efficiency: Save time and resources with our efficient automated process.
• Flexibility: Adapt the system to your needs, whether you’re dealing with tweets, posts, meta-descriptions, or detailed product descriptions.

🌐 Unlimited Possibilities: Once the model is created, it can handle an unlimited number of ads, regardless of their complexity. Whether it’s a short tweet or a detailed product description, our system can handle it all!

Ready to streamline your content classification process? Try our system today and experience the future of content management!

4.5 - Talent matching

Classify personnel skills according to your taxonomy to improve the profile creation and achieve a better match with job offerings and job postings.

5 - Prices

Please contact us for special a quote . Offerings plans start form 100E. Free trial also available.

The EZservices are billed based on the following factors:

• The number of created models.
• The amount of data to be classified.

You have the option to choose from pre-paid packages or monthly plans.

For pre-paid packages, please note that they must be used within 12 months.