erickfmm/Tesis_UChile

Tesis UChile

erickfmm/Tesis_UChile is a collection of thesis metadata from the University of Chile repository (repositorio.uchile.cl). The data was obtained by scraping public pages around August 2024 (approximately), and then normalized into a row-based CSV plus grouped JSON exports.

What is in this dataset?

The main file is train.csv. Each row represents one metadata value associated with a source record. Rows are linked by the origen field, which acts as the record identifier or source URL.

The repository also includes JSON and JSONL exports generated from train.csv, where all rows that share the same origen are grouped into a single document.

This dataset mainly contains bibliographic and repository metadata, for example:

• thesis title
• author
• advisor
• publisher / faculty / school
• issue or accession dates
• language
• subjects
• rights / access fields
• repository links and file links

Most keys follow Dublin Core naming conventions such as dc.title, dc.subject, dc.identifier.uri, and dcterms.accessRights.

Files

File	Description	Link
train.csv	Row-level metadata table used as the base dataset. On Hugging Face, this is exposed as the train split.	Download
tesis_uchile_pretty.json	Grouped JSON array, pretty-printed for easier inspection.	Download
tesis_uchile_min.json	Grouped JSON array, compact version of the same data.	Download
tesis.uchile.jsonl	Grouped JSON Lines export, one record per line, convenient for streaming and incremental processing.	Download

CSV schema

The processing script expects at least the following columns in train.csv:

Column	Meaning
origen	Source identifier for a thesis record, typically the repository URL used to group rows.
DC	Metadata field name, usually a Dublin Core-style key such as dc.title or dc.subject.
nombre	Original scraped field name. A special case maps filelink to extra.file.link when DC is missing.
value	Metadata value. Empty values are skipped when building the JSON exports.
lang	Optional language tag for the value, for example es_CL.

JSON / JSONL structure

The JSON exports are built by grouping all rows with the same origen.

Rules used in the export:

• each grouped record gets _id = origen
• scalar values are stored as {"value": "..."}
• when a language is available, values become {"value": "...", "lang": "..."}
• repeated metadata fields become lists
• if DC is missing and nombre == "filelink", the field is stored as extra.file.link

Example grouped record:

{
  "_id": "https://repositorio.uchile.cl/handle/2250/100054",
  "dc.contributor.advisor": {
    "value": "Cárdenas, Juan",
    "lang": "es_CL"
  },
  "dc.contributor.author": {
    "value": "Gallego, Francisco",
    "lang": "es_CL"
  },
  "dc.contributor.editor": [
    {
      "value": "Facultad de Arquitectura y Urbanismo",
      "lang": "es_CL"
    },
    {
      "value": "Escuela de Arquitectura",
      "lang": "es_CL"
    }
  ],
  "dc.identifier.uri": {
    "value": "https://repositorio.uchile.cl/handle/2250/100054"
  },
  "extra.file.link": {
    "value": "..."
  }
}

Example code

Load the CSV with pandas

import pandas as pd

df = pd.read_csv(
    "hf://datasets/erickfmm/Tesis_UChile/train.csv",
    encoding="utf-8",
)

print(df.head())
print(df.columns.tolist())

Load the dataset with datasets

from datasets import load_dataset

dataset = load_dataset("erickfmm/Tesis_UChile")
print(dataset)
print(dataset["train"][0])

Stream the JSONL export

import json
import urllib.request

url = "https://huggingface.co/datasets/erickfmm/Tesis_UChile/resolve/main/tesis.uchile.jsonl"

with urllib.request.urlopen(url) as response:
    for i, line in enumerate(response):
        record = json.loads(line)
        title = record.get("dc.title", {})
        print(record["_id"], title)
        if i == 2:
            break

Read the compact JSON export locally

import json

with open("tesis_uchile_min.json", "r", encoding="utf-8") as fh:
    data = json.load(fh)

print(len(data))
print(data[0]["_id"])

Source and collection notes

• Source website: public pages from the University of Chile repository (repositorio.uchile.cl)
• Collection method: scraping
• Collection period: around August 2024 (approximate)
• Contents: public metadata records and derived grouped exports

License

No license is provided for this dataset in this repository, or the license is unknown.

If you plan to reuse the data, especially beyond research or indexing purposes, please verify the terms of use of the original source and the rights associated with individual records.

Limitations

• scraped metadata can contain inconsistencies, missing fields, duplicates, or formatting issues
• language tags are not guaranteed to exist for every value
• the JSON exports are derived from train.csv, so updates to the CSV should be regenerated in the exported formats