How do datasets work

01 · How to build a dataset

IN FIVE STEPS

1. 01

   Name the group

   Give the group a name you will recognise — it travels into the export as a label. For example, "Pension reform_2018"

   NEW GROUP

   Duplicate

   01 Condition

   NOT

   All fields

   All of these words

   Add criterion

   Add condition
2. 02

   Add your conditions and criterions

   Inside the group, add your conditions and criterions. You can write the criterions both in Russian and English. An English query will match a material if we hold a translation of it, as well as a material that was originally written in English.

   ANDOR

   All conditions must be met

   01 Condition

   NOT

   ANDOR

   All of the criteria must be met

   All fields

   This exact phrase

   AND

   All fields

   Any of these words

   Add criterion

   AND

   02 Condition

   NOT

   All fields

   All of these words

   Add criterion

   Add condition

   A material enters the group if it contains both the exact phrase "пенсионный возраст" (age of retirement), and at least one of "митинг", "протест", "акция" (rally/protest/demonstration), and does not contain "Франция" (France).

   The NOT toggle turns a condition into an exclusion: materials matching it are removed from the group. It earns its place here because France saw its own pension protests in 2018 and 2019, covered heavily by Russian media; without the exclusion they would enter the group and distort the statistics.

   WORTH KNOWING

   Avoid single high-frequency words. The word пенсия (pension) alone returns tens of thousands of materials across all years, and the resulting dataset is of no use to anyone. What works is exact phrases (пенсионная реформа/pension reform) and combinations of conditions, as above.

3. 03

   Narrow the group with filters

   Filters apply to the group as a whole: sources, authors, source types (web, print, video) and publication period. An empty filter means no restriction.

   For the pension reform group, for instance, it makes sense to set the period from 01.01.2018 to 31.12.2019. Outside those dates there are less relevant matches, while the volume of the export grows noticeably.

   SOURCES (164)AUTHORS (285)TYPES (3)DATES (2018—2019)

   Hide filters and dates(2)

   DATE RANGE

   From:

   //

   Until:

   //

   WORTH KNOWING

   A dataset can be built from filters alone

   If what you need is every material from one or several sources over a given period, leave the conditions empty and set only sources and dates. This is the right approach when you intend to search the dataset yourself, with your own tools: we hand you the entire slice, and what you do with it afterwards is up to you.
   The same holds for the other filters. A group with no conditions but a selected author returns everything that author published in the archive.

   Selected source filters:

   Meduza

   +1 source

   Meduza

   Selected date filters:

   From 06/01/2018

   To 10/31/2018

   2

   1

   From 06/01/2018To 10/31/2018

   DELETE GROUPSAVE GROUP

   WORTH KNOWING

   Maximum size of a single export — 500,000 materials.

   WORTH KNOWING

   Filters narrow, they do not accumulate, and they can rule each other out.

   If you select only web outlets under sources and then only print under source types, no material satisfies both, and the group comes back empty. The same happens when the author you selected never published in the sources you selected, or when the period does not overlap with the years a source existed.
   If a group returns nothing, check the filters before you check the conditions.

4. 04

   Add the remaining groups

   If you are comparing several topics, periods or regions, make a separate group for each. One request produces one dataset, but it may hold as many groups as you need. A material matching several groups arrives once and carries all the labels that apply to it.

   BUILD A DATASET

   Collected 12,798 materials out of 500,000 possible

   Imported from advanced search

   Group name

   6,399materials

   Edit group

   Group name

   6,399materials

   Edit group

   ADD GROUP
5. 05

   Check what you have built

   Before sending, read the resulting formulation of your query — we render it in plain language beneath each group. If the formulation describes something other than what you had in mind, adjust the conditions.

   Resulting query:

   ALL of the following are true:01.The title or text contains the exact phrase “пенсионный возраст”and contains at least one of these words: “митинг протест акция”02.EXCLUDE — the title or text contains the word “Франция”

   Selected source filters:

   Meduza

   +1 source

   Meduza

   Selected date filters:

   From 01/01/2018

   To 12/31/2019

   2

   1

   From 01/01/2018To 12/31/2019

   DELETE GROUPSAVE GROUP

02 · A worked example

DATASET: «ПЕНСИОННАЯ РЕФОРМА 2018»

Group 1:

NEW GROUP

Duplicate

ANDOR

All conditions must be met

01 Condition

NOT

ANDOR

All of the criteria must be met

All fields

This exact phrase

AND

All fields

Any of these words

Add criterion

AND

02 Condition

NOT

All fields

All of these words

Add criterion

Add condition

SOURCES (1)AUTHORS (285)TYPES (3)DATES (2018—2019)

Hide filters and dates(3)

DATE RANGE

From:

//

Until:

//

Selected source filters:

Meduza

+1 source

Meduza

Selected date filters:

From 01/01/2018

To 12/31/2019

2

1

From 01/01/2018To 12/31/2019

DELETE GROUPSAVE GROUP

03 · Download a sample

Look at the data before you send a request

04 · What you get and when

05 · Questions