A knowledge graph querying framework
diff --git a/pages/research.md b/pages/research.md
index 046d2646a..7673ee1cf 100644
--- a/pages/research.md
+++ b/pages/research.md
@@ -10,7 +10,7 @@ as a framework for researching query execution over decentralized knowledge grap
Since Comunica is open-source, anyone is free to use it to perform any research they want.
If you are interested in **collaborating**, or working on Comunica as a **Bachelor/Master/PhD student**,
-feel free to [contact us](ask/#email).
+feel free to [contact us](/ask/#email).
We always have interesting projects to offer on both applied development and fundamental research.
## Cite
@@ -48,7 +48,7 @@ The following publications make significant use of Comunica:
* [**SAD Generator: eating our own dog food to generate KGs and websites for academic events**. Heyvaert, P., Chaves-Fraga, D., Priyatna, F., Sequeda, J., & Dimou, A. (2019, June).](https://link.springer.com/chapter/10.1007/978-3-030-32327-1_19)
* [**Versioned Querying with OSTRICH and Comunica in MOCHA 2018**. Taelman, R., Vander Sande, M., & Verborgh, R. (2018, June)](https://biblio.ugent.be/publication/8566999/file/8567001.pdf)
-Also using Comunica in our work? [Let us know](ask/#email) so we can add a reference to this list.
+Also using Comunica in our work? [Let us know](/ask/#email) so we can add a reference to this list.
## Tutorials
diff --git a/pages/research/amf.md b/pages/research/amf.md
index 776506b52..9b25db729 100644
--- a/pages/research/amf.md
+++ b/pages/research/amf.md
@@ -29,7 +29,7 @@ Due to the reduction of HTTP requests, complete results come in earlier.
In some cases, the first result can be delayed.
-
+
### Caching significantly speeds up query execution
@@ -37,7 +37,7 @@ In some cases, the first result can be delayed.
An HTTP cache like NGINX achieves the best results, but additionally caching AMF filters server-side is not worth the effort.
-
+
### Extreme false-positive probabilities slow down query execution
@@ -45,7 +45,7 @@ An HTTP cache like NGINX achieves the best results, but additionally caching AMF
On average, a false-positive probability of 1/64 leads to the lowest overall query evaluation times for this experiment.
-
+
## 3. Recommendations for data publishers