Overview

UiPath provides a number of machine learning capabilities out-of-the-box on UiPath® AI Center. A notable example is Document Understanding^TM. In addition, UiPath built and open-source models (serving-only and retrainable) are continuously added to AI Center.

Note: When creating an ML Package in AI Center, it cannot be named using any python reserved keyword, such as class, break, from, finally, global, None, etc. Make sure to choose another name. The listed examples are not complete since package name is used for class <pkg-name> and import <pck-name>.

Note: These models are supported as-is and are not actively worked on to improve model performance. Since the code is open-sourced, it is fully accessible for optimizations. Bugs with respect to usage on the UiPath platform are investigated and fixed on a best effort basis with no SLA guarantees. These models are not recommended for use in critical production workflows.

The following packages are available in platform today :

Model	Category	Type	Availability
Image Classification	UiPath Image Analysis	Custom Training	Preview
Signature Comparison	UiPath Image Analysis	Pre Trained	General Availability
Custom Named Entity Recognition	UiPath Language Analysis	Custom Training	General Availability
Light Text Classification	UiPath Language Analysis	Custom Training	General Availability
Multilingual Text Classification	UiPath Language Analysis	Custom Training	General Availability
Semantic Similarity	UiPath Language Analysis	Pre Trained	Preview
Multilabel Text Classification	UiPath Language Analysis	Custom Training	Preview
TM Analyzer Model	UiPath Task Mining	Custom Training	General Availability
Image Moderation	Open-Source Packages - Image Analysis	Pre Trained	N/A
Object Detection	Open-Source Packages - Image Analysis	Pre Trained and Custom Training	N/A
English Text Classification	Open-Source Packages - Language Analysis	Custom Training	N/A
French Text Classification	Open-Source Packages - Language Analysis	Custom Training	N/A
Japanese Text Classification	Open-Source Packages - Language Analysis	Custom Training	N/A
Language Detection	Open-Source Packages - Language Analysis	Pre Trained	N/A
Named Entity Recognition	Open-Source Packages - Language Analysis	Pre Trained	N/A
Sentiment Analysis	Open-Source Packages - Language Analysis	Pre Trained	N/A
Text Classification	Open-Source Packages - Language Analysis	Custom Training	N/A
Question Answering	Open-Source Packages - Language Comprehension	Pre Trained	N/A
Semantic Similarity	Open-Source Packages - Language Comprehension	Pre Trained	N/A
Text Summarization	Open-Source Packages - Language Comprehension	Pre Trained	N/A
English To French Translation	Open-Source Packages - Language Translation	Pre Trained	N/A
English To German Translation	Open-Source Packages - Language Translation	Pre Trained	N/A
English To Russian Translation	Open-Source Packages - Language Translation	Pre Trained	N/A
German To English Translation	Open-Source Packages - Language Translation	Pre Trained	N/A
MultilingualTranslator	Open-Source Packages - Language Translation	Pre Trained	N/A
Russian To English Translation	Open-Source Packages - Language Translation	Pre Trained	N/A
TPOT Tabular Classification	Open-Source Packages - Tabular Data	Custom Training	N/A
TPOT Tabular Regression	Open-Source Packages - Tabular Data	Custom Training	N/A
XGBoost Tabular Classification	Open-Source Packages - Tabular Data	Custom Training	N/A
XGBoost Tabular Regression	Open-Source Packages - Tabular Data	Custom Training	N/A

Note: For Document Understanding models, check the Document Understanding guide.

Ready-to-deploy

Example packages that can be immediately deployed and added to a RPA workflow, more can be found in the product

Image Moderation

This is a model for image content moderation based on a deep learning architecture commonly referred to as Inception V3. Given an image, the model will output one of four classes 'explicit', 'explicit-drawing', 'neutral', and 'pornographic' together with a normalized confidence score for each class probability.

It is based on the paper 'Rethinking the Inception Architecture for Computer Vision' by Szegedy et al which was open-sourced by Google.

Sentiment Analysis

This model predicts the sentiment of a text in the English Language. It was open-sourced by Facebook Research. Possible predictions are one of "Very Negative", "Negative", "Neutral", "Positive", "Very Positive". The model was trained on Amazon product review data thus, the model predictions may have some unexpected results for different data distributions. A common use case is to route unstructured language content (e.g. emails) based on the sentiment of the text.

It is based on the research paper "Bag of Tricks for Efficient Text Classification" by Joulin, et al.

Question Answering

This model predicts the answer to a question of a text in the English Language based on some paragraph context. It was open-sourced by ONNX. A common use case is in KYC or processing financial reports where a common question can be applied to a standard set of semi-structured documents. It is based on the state-of-the-art BERT (Bidirectional Encoder Representations from Transformers). The model applies Transformers, a popular attention model, to language modeling to produce an encoding of the input and then trains on the task of question answering.

It is based on the research paper “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding”.

Language Identification

This model predicts the language of a text input. Possible predictions are one of the following 176 languages:

Languages

af als am an ar arz as ast av az azb ba bar bcl be bg bh bn bo bpy br bs bxr ca cbk ce ceb ckb co cs cv cy da de diq dsb dty dv el eml en eo es et eu fa fi fr frr fy ga gd gl gn gom gu gv he hi hif hr hsb ht hu hy ia id ie ilo io is it ja jbo jv ka kk km kn ko krc ku kv kw ky la lb lez li lmo lo lrc lt lv mai mg mhr min mk ml mn mr mrj ms mt mwl my myv mzn nah nap nds ne new nl nn no oc or os pa pam pfl pl pms pnb ps pt qu rm ro ru rue sa sah sc scn sco sd sh si sk sl so sq sr su sv sw ta te tg th tk tl tr tt tyv ug uk ur uz vec vep vi vls vo wa war wuu xal xmf yi yo yue zh

It was open-sourced by Facebook Research. The model was trained on data from Wikipedia, Tatoeba, and SETimes used under the Creative Commons Attribution-Share-Alike License 3.0. A common use case is to route unstructured language content (e.g. emails) to an appropriate responder based on the language of the text.

It is based on the research paper "Bag of Tricks for Efficient Text Classification" by Joulin, et al.

Multilingual Translator

This model delivers translation directly between any pair of 200+ languages. You can find the full list of languages and the corresponding code to use each of them here.

This is the HuggingFace integration of No Language Left Behind model open sourced by Meta AI Research. The model was published under the following license: License.

Input description

The model input is JSON with three keys:

• text: the text to be translated.
• from_lang: the language code of the text to be translated.
• to_lang: the language code of the targeted text.

For example:

{"text" : "UN Chief says there is no military solution in Syria", "from_lang" : "eng_Latn", "to_lang" : "fra_Latn" }"

Output description

The model returns the translated text in targeted language:

"Le chef de l'ONU dit qu'il n'y a pas de solution militaire en Syrie"

English To French

This is a Sequence-to-Sequence machine translation model that translates English to French. It was open-sourced by Facebook AI Research (FAIR).

It is based on the paper "Convolutional Sequence to Sequence Learning" by Gehring, et al.

English To German

This is a Sequence-to-Sequence machine translation model that translates English to German. It was open-sourced by Facebook AI Research (FAIR).

It is based on the paper "Facebook FAIR's WMT19 News Translation Submission" by Ng, et al.

German To English

This is a Sequence-to-Sequence machine translation model that translates English to Russian. It was open-sourced by Facebook AI Research (FAIR).

It is based on the paper "Facebook FAIR's WMT19 News Translation Submission" by Ng, et al.

English To Russian

This is a Sequence-to-Sequence machine translation model that translates English to Russian. It was open-sourced by Facebook AI Research (FAIR).

It is based on the paper "Facebook FAIR's WMT19 News Translation Submission" by Ng, et al.

Russian To English

This is a Sequence-to-Sequence machine translation model that translates English to Russian. It was open-sourced by Facebook AI Research (FAIR).

It is based on the paper "Facebook FAIR's WMT19 News Translation Submission" by Ng, et al.

NamedEntityRecognition

This model returns a list of entities recognized in text. The 18 types of named entities recognized use the same output class as in OntoNotes5 which is commonly used for benchmarking this task in academia. The model is based on the paper 'Approaching nested named entity recognition with parallel LSTM-CRFs' by Borchmann et al, 2018.

The 18 classes are the following:

Entity	Description
PERSON	People, including fictional.
NORP	Nationalities or religious or political groups.
FAC	Buildings, airports, highways, bridges, etc.
ORG	Companies, agencies, institutions, etc.
GPE	Countries, cities, states.
LOC	Non-GPE locations, mountain ranges, bodies of water.
PRODUCT	Objects, vehicles, foods, etc. (Not services.)
EVENT	Named hurricanes, battles, wars, sports events, etc.
WORK_OF_ART	Titles of books, songs, etc.
LAW	Named documents made into laws.
LANGUAGE	Any named language.
DATE	Absolute or relative dates or periods.
TIME	Times smaller than a day.
PERCENT	Percentage, including ”%“.
MONEY	Monetary values, including unit.
QUANTITY	Measurements, as of weight or distance.
ORDINAL	“first”, “second”, etc.
CARDINAL	Numerals that do not fall under another type.

Re-trainable

Example packages that can be trained by adding data to AI Center storage and starting a pipeline, more models can be found in the product.

English Text Classification

This is a generic, re-trainable model for English text classification. Common use cases are email classification, service ticket classification, custom sentiment analysis among others. See English Text Classification for more details.

French Text Classification

This is a generic, re-trainable model for French text classification. Common use cases are email classification, service ticket classification, custom sentiment analysis among others. See French Text Classification for more details.

Multi Lingual Text Classification

This is the preview version of a generic, retrainable model for text classification. It supports the top 100 Wikipedia languages listed here. This ML Package must be trained, and if deployed without training first, the deployment will fail with an error stating that the model is not trained. It is based on BERT, a self-supervised method for pretraining natural language processing systems. A GPU is recommended especially during training. A GPU delivers ~5-10x improvement in speed.

Custom Named Entity Recognition

This preview model allows you to bring your own dataset tagged with entities you want to extract. The training and evaluation datasets need to be in CoNLL format.

Tabular Classification AutoML - TPOT

This is a generic, re-trainable model for tabular (e.g. csv, excel) data classification. That is, given a table of columns and a target column, it will find a model for that data. See TPOT AutoML Classification for more details.

Tabular Classification - TPOT XGBoost

This is a generic, re-trainable model for tabular (e.g. csv, excel) data classification. That is, given a table of columns and a target column, it will find a model (based on XGBoost) for that data. See TPOT XGBoost Classification,