Develop Text Classification and/or Clustering Algorithms in Python

We require assistance on the following tasks. Please contact us directly to describe how you would solve them. Russian language skills may be necessary. 1) Task: Develop/employ a text-classification algorithm in Python or R that classifies items as one of several thousand 10-digit product codes using a descriptive text field of roughly 300 characters in UTF-8 (Russian / Cyrillic). Description: We have a database of several million textual descriptions of products that have been entered by humans. Each entry is connected to a 10 digit product code, but the same product code can be used for multiple differing textual entries. We require a text-classification algorithm that probabilistically classifies a document that can then be applied to another dataset (see task 2). This task requires tokenizing, stemming, and removing stop words, and therefore you may need to know Russian or to use available NLTK packages. Similarly, several different algorithms may need to be used to improve precision. Output: Python scripts/algorithm(s) classifying documents into 10-digit product codes that can be used in task 2. 2) Task: Use the classification algorithm in (1) to classify textual entries in a second dataset. Description: Once the clean list has been created, employ a machine learning algorithm to assign the 10 digit codes to a target dataset of over 60 million textual product descriptions in UTF-8 (Russian / Cyrillic). Not all entries will have sufficient information to be classified ('leftovers') and should be marked so. For example, this could be done if no classification has a probability above some threshold. Also, the dataset in (1) only contains examples of a subset of the items in the second dataset, but we will be able to estimate which items these are. Output: Second dataset of 60 million entries are matched to 10 digit product codes. 3) Task: For the 'leftovers' of (2), develop/employ a text clustering algorithm that groups entries in k subclasses Description: We will provide you a higher-level grouping variable for the 'leftovers' and a number k that designates how many we clusters need within each grouping. Your task will be to use a text clustering algorithm to create k amount of 'clusters' within the higher groups for the 'leftovers'. Output: A unique variable designating cluster membership for each item in the 'leftovers' (those without 10 digit product codes from step 2).